Panoramic video generation has attracted growing attention due to its applications in virtual reality and immersive media. However, existing methods lack explicit motion control and struggle to generate scenes with large and complex motions. We propose PanFlow a novel approach that exploits the spherical nature of panoramas to decouple the highly dynamic camera rotation from the input optical flow condition, enabling more precise control over large and dynamic motions. We further introduce a spherical noise warping strategy to promote loop consistency in motion across panorama boundaries. To support effective training, we curate a large-scale, motion-rich panoramic video dataset with frame-level pose and flow annotations. We also showcase the effectiveness of our method in various applications, including motion transfer and video editing. Extensive experiments demonstrate that PanFlow significantly outperforms prior methods in motion fidelity, visual quality, and temporal coherence.

Semi-supervised learning in intrusion detection systems (IDS) faces three major challenges: the scarcity of labeled samples, class imbalance, and distribution divergence between labeled and unlabeled data. Moreover, the third challenge may lead to an extreme situation where labeled data fail to cover all categories. To address these issues, we develop a heuristic factors based semi-supervised model named HF-IDS. Specifically, we employ symbolic regression, a novel feature engineering technique, to generate class-indicating factors. These factors guide a multi-level clustering for pseudo-label addition to some unlabeled data. For classification, we construct an Ensemble of Graph Neural Networks (E-GNNs). Different from common ensemble learning methods, each of the GNN classifiers is equipped with a unique graph filter constructed by the factors. Through these filters, we topologically reweight different classes to enhance the model’s effectiveness in handling with class imbalance problem. The model is evaluated in both multi-class and binary classification settings, with the binary case simulating the extreme missing-category scenario. Experiments on NSL-KDD and CICIDS-2017 datasets show that HF-IDS consistently outperforms state-of-the-art baselines in accuracy, precision, recall, and F1_score.

Overcoming class imbalance is a critical challenge for graph-based semi-supervised classification methods. In this letter, we address this issue from the perspective of graph filtering and propose a novel adaptive graph filter. By introducing learnable thresholds into the adjacency matrix, the filter enables dynamic suppression of majority-class bias during label propagation through the incorporation of discontinuities. Additionally, we develop a modified Fruit Fly Optimization Algorithm (m-FOA) to optimize the filter's coefficients, which achieves lower loss and faster convergence compared to other heuristic algorithms. To evaluate the effectiveness of our approach, we conduct a Monte Carlo simulation on a real-world dataset. The results demonstrate that our method outperforms the baseline methods in both classification accuracy and efficiency when handling class imbalance. We note that the model's scalability to very large graphs is limited and the solving procedure can be time-consuming due to the dense construction of the filter.

How can one efficiently generate high-quality, wide-scope 3D scenes from arbitrary single images? Existing methods suffer several drawbacks, such as requiring multi-view data, time-consuming per-scene optimization, distorted geometry in occluded areas, and low visual quality in backgrounds. Our novel 3D scene reconstruction pipeline overcomes these limitations to tackle the aforesaid challenge. Specifically, we introduce a large-scale reconstruction model that leverages latents from a video diffusion model to predict 3D Gaussian Splattings of scenes in a feed-forward manner. The video diffusion model is designed to create videos precisely following specified camera trajectories, allowing it to generate compressed video latents that encode multi-view information while maintaining 3D consistency. We train the 3D reconstruction model to operate on the video latent space with a progressive learning strategy, enabling the efficient generation of high-quality, wide-scope, and generic 3D scenes. Extensive evaluations across various datasets affirm that our model significantly outperforms existing single-view 3D scene generation methods, especially with out-of-domain images. Thus, we demonstrate for the first time that a 3D reconstruction model can effectively be built upon the latent space of a diffusion model in order to realize efficient 3D scene generation. Project page: https://snap-research.github.io/wonderland/

Dataset distillation has demonstrated strong performance on simple datasets like CIFAR, MNIST, and TinyImageNet but struggles to achieve similar results in more complex scenarios. In this paper, we propose EDF (emphasizes the discriminative features), a dataset distillation method that enhances key discriminative regions in synthetic images using Grad-CAM activation maps. Our approach is inspired by a key observation: in simple datasets, high-activation areas typically occupy most of the image, whereas in complex scenarios, the size of these areas is much smaller. Unlike previous methods that treat all pixels equally when synthesizing images, EDF uses Grad-CAM activation maps to enhance high-activation areas. From a supervision perspective, we downplay supervision signals produced by lower trajectory-matching losses, as they contain common patterns. Additionally, to help the DD community better explore complex scenarios, we build the Complex Dataset Distillation (Comp-DD) benchmark by meticulously selecting sixteen subsets, eight easy and eight hard, from ImageNet-1K. In particular, EDF consistently outperforms SOTA results in complex scenarios, such as ImageNet-1K subsets. Hopefully, more researchers will be inspired and encouraged to improve the practicality and efficacy of DD. Our code and benchmark have been made public at NUS-HPC-AI-Lab/EDF.

Early diagnosis and treatment of polyps during colonoscopy are essential for reducing the incidence and mortality of Colorectal Cancer (CRC). However, the variability in polyp characteristics and the presence of artifacts in colonoscopy images and videos pose significant challenges for accurate and efficient polyp detection and segmentation. This paper presents a novel approach to polyp segmentation by integrating the Segment Anything Model (SAM 2) with the YOLOv8 model. Our method leverages YOLOv8’s bounding box predictions to autonomously generate input prompts for SAM 2, thereby reducing the need for manual annotations. We conducted exhaustive tests on five benchmark colonoscopy image datasets and two colonoscopy video datasets, demonstrating that our method exceeds state-of-the-art models in both image and video segmentation tasks. Notably, our approach achieves high segmentation accuracy using only bounding box annotations, significantly reducing annotation time and effort. This advancement holds promise for enhancing the efficiency and scalability of polyp detection in clinical settings https://github.com/sajjad-sh33/YOLO_SAM2.

CLIP exhibits strong visual-textual alignment but struggle with open-vocabulary segmentation due to poor localization. Prior methods enhance spatial coherence by modifying intermediate attention. But, this coherence isn't consistently propagated to the final output due to subsequent operations such as projections. Additionally, intermediate attention lacks direct interaction with text representations, such semantic discrepancy limits the full potential of CLIP. In this work, we propose a training-free, feedback-driven self-adaptive framework that adapts output-based patch-level correspondences back to the intermediate attention. The output predictions, being the culmination of the model's processing, encapsulate the most comprehensive visual and textual semantics about each patch. Our approach enhances semantic consistency between internal representations and final predictions by leveraging the model's outputs as a stronger spatial coherence prior. We design key modules, including attention isolation, confidence-based pruning for sparse adaptation, and adaptation ensemble, to effectively feedback the output coherence cues. Our method functions as a plug-in module, seamlessly integrating into four state-of-the-art approaches with three backbones (ViT-B, ViT-L, ViT-H). We further validate our framework across multiple attention types (Q-K, self-self, and Proxy augmented with MAE, SAM, and DINO). Our approach consistently improves their performance across eight benchmarks.

Vision Language Models (VLMs) have rapidly advanced in integrating visual and textual reasoning, powering applications across high-resolution image understanding, long-video analysis, and multi-turn conversation. However, their scalability remains limited by the growing number of visual tokens that dominate inference latency. We present SparseVILA, a new paradigm for efficient VLM inference that decouples visual sparsity across the prefilling and decoding stages. SparseVILA distributes sparsity across stages by pruning redundant visual tokens during prefill and retrieving only query-relevant tokens during decoding. This decoupled design matches leading prefill pruning methods while preserving multi-turn fidelity by retaining most of the visual cache so that query-aware tokens can be retrieved at each conversation round. Built on an AWQ-optimized inference pipeline, SparseVILA achieves up to 4.0 times faster prefilling, 2.5 times faster decoding, and an overall 2.6 times end-to-end speedup on long-context video tasks -- while improving accuracy on document-understanding and reasoning tasks. By decoupling query-agnostic pruning and query-aware retrieval, SparseVILA establishes a new direction for efficient multimodal inference, offering a training-free, architecture-agnostic framework for accelerating large VLMs without sacrificing capability.

Using massive datasets, foundation models are large-scale, pre-trained models that perform a wide range of tasks. These models have shown consistently improved results with the introduction of new methods. It is crucial to analyze how these trends impact the medical field and determine whether these advancements can drive meaningful change. This study investigates the application of recent state-of-the-art foundation models—DINOv2, MAE, VMamba, CoCa, SAM2, and AIMv2—for medical image classification. We explore their effectiveness on datasets including CBIS-DDSM for mammography, ISIC2019 for skin lesions, APTOS2019 for diabetic retinopathy, and CHEXPERT for chest radiographs. By fine-tuning these models and evaluating their configurations, we aim to understand the potential of these advancements in medical image classification. The results indicate that these advanced models significantly enhance classification outcomes, demonstrating robust performance despite limited labeled data. Based on our results, AIMv2, DI-NOv2, and SAM2 models outperformed others, demonstrating that progress in natural domain training has positively impacted the medical domain and improved classification outcomes. Our code is publicly available at https://github.com/sajjad-sh33/Medical-Transfer-Learning.

Few-shot Test-Time Domain Adaptation focuses on adapting a model at test time to a specific domain using only a few unlabeled examples, addressing domain shift. Prior methods leverage CLIP's strong out-of-distribution (OOD) abilities by generating domain-specific prompts to guide its generalized, frozen features. However, since downstream datasets are not explicitly seen by CLIP, solely depending on the feature space knowledge is constrained by CLIP's prior knowledge. Notably, when using a less robust backbone like ViT-B/16, performance significantly drops on challenging real-world benchmarks. Departing from the state-of-the-art of inheriting the intrinsic OOD capability of CLIP, this work introduces learning directly on the input space to complement the dataset-specific knowledge for frozen CLIP. Specifically, an independent side branch is attached in parallel with CLIP and enforced to learn exclusive knowledge via revert attention. To better capture the dataset-specific label semantics for downstream adaptation, we propose to enhance the inter-dispersion among text features via greedy text ensemble and refinement. The text and visual features are then progressively fused in a domain-aware manner by a generated domain prompt to adapt toward a specific domain. Extensive experiments show our method's superiority on 5 large-scale benchmarks (WILDS and DomainNet), notably improving over smaller networks like ViT-B/16 with gains of \textbf{+5.1} in F1 for iWildCam and \textbf{+3.1\%} in WC Acc for FMoW. \href{https://github.com/chi-chi-zx/L2C}{Our Code: L2C}

Fine-tuning LLMs is both computationally and memory-intensive. While parameter-efficient fine-tuning methods, such as QLoRA and DoRA, reduce the number of trainable parameters and lower memory usage, they do not decrease computational cost. In some cases, they may even slow down fine-tuning. In this paper, we introduce SparseLoRA, a method that accelerates LLM fine-tuning through contextual sparsity. We propose a lightweight, training-free SVD sparsity estimator that dynamically selects a sparse subset of weights for loss and gradient computation. Also, we systematically analyze and address sensitivity across layers, tokens, and training steps. Our experimental results show that SparseLoRA reduces computational cost by up to $2.2\times$ and a measured speedup of up to $1.6\times$ while maintaining accuracy across various downstream tasks, including commonsense and arithmetic reasoning, code generation, and instruction following.

With the rapid advancement and widespread adoption of VR/AR technologies, there is a growing demand for the creation of high-quality, immersive dynamic scenes. However, existing generation works predominantly concentrate on the creation of static scenes or narrow perspective-view dynamic scenes, falling short of delivering a truly 360-degree immersive experience from any viewpoint. In this paper, we introduce TiP4GEN, an advanced text-to-dynamic panorama scene generation framework that enables fine-grained content control and synthesizes motion-rich, geometry-consistent panoramic 4D scenes. TiP4GEN integrates panorama video generation and dynamic scene reconstruction to create 360-degree immersive virtual environments. For video generation, we introduce a Dual-branch Generation Model consisting of a panorama branch and a perspective branch, responsible for global and local view generation, respectively. A bidirectional cross-attention mechanism facilitates comprehensive information exchange between the branches. For scene reconstruction, we propose a Geometry-aligned Reconstruction Model based on 3D Gaussian Splatting. By aligning spatial-temporal point clouds using metric depth maps and initializing scene cameras with estimated poses, our method ensures geometric consistency and temporal coherence for the reconstructed scenes. Extensive experiments demonstrate the effectiveness of our proposed designs and the superiority of TiP4GEN in generating visually compelling and motion-coherent dynamic panoramic scenes.

Next-generation wireless networks demand scalable and low-latency distributed learning frameworks that can operate across diverse communication environments. This paper proposes CFVLC, a novel large-scale Federated Learning (FL) framework over a resource-constrained wireless network, involving outdoor users connected through Cell-Free massive MIMO (CF-mMIMO) system, and indoor users distributed across multiple environments supported by Visible Light Communication (VLC) technology. To address the extreme system heterogeneity from different environments, CFVLC introduces an optimization problem that jointly performs device selection and resource allocation to minimize training latency across the network. We propose a two-stage heuristic solution that intelligently balances user participation across environments with varying resources and channel conditions. Extensive simulations demonstrate that our approach significantly improves training latency and model performance compared to conventional schemes, highlighting the benefits of integrating outdoor CF-mMIMO and indoor VLC technologies for large-scale FL.

Despite the recent remarkable achievement in gaze estimation, efficient and accurate personalization of gaze estimation without labels is a practical problem but rarely touched on in the literature. To achieve efficient personalization, we take inspiration from the recent advances in Natural Language Processing (NLP) by updating a negligible number of parameters, "prompts", at the test time. Specifically, the prompt is additionally attached without perturbing original network and can contain less than 1% of a ResNet-18's parameters. Our experiments show high efficiency of the prompt tuning approach. The proposed one can be 10 times faster in terms of adaptation speed than the methods compared. However, it is non-trivial to update the prompt for personalized gaze estimation without labels. At the test time, it is essential to ensure that the minimizing of particular unsupervised loss leads to the goals of minimizing gaze estimation error. To address this difficulty, we propose to meta-learn the prompt to ensure that its updates align with the goal. Our experiments show that the meta-learned prompt can be effectively adapted even with a simple symmetry loss. In addition, we experiment on four cross-dataset validations to show the remarkable advantages of the proposed method.

Test-time domain adaptation aims to adapt the model trained on source domains to unseen target domains using a few unlabeled images. Emerging research has shown that the label and domain information is separately embedded in the weight matrix and batch normalization (BN) layer. Previous works normally update the whole network naively without explicitly decoupling the knowledge between label and domain. As a result, it leads to knowledge interference and defective distribution adaptation. In this work, we propose to reduce such learning interference and elevate the domain knowledge learning by only manipulating the BN layer. However, the normalization step in BN is intrinsically unstable when the statistics are re-estimated from a few samples. We find that ambiguities can be greatly reduced when only updating the two affine parameters in BN while keeping the source domain statistics. To further enhance the domain knowledge extraction from unlabeled data, we construct an auxiliary branch with label-independent self-supervised learning (SSL) to provide supervision. Moreover, we propose a bi-level optimization based on meta-learning to enforce the alignment of two learning objectives of auxiliary and main branches. The goal is to use the auxiliary branch to adapt the domain and benefit main task for subsequent inference. Our method keeps the same computational cost at inference as the auxiliary branch can be thoroughly discarded after adaptation. Extensive experiments show that our method outperforms the prior works on five WILDS real-world domain shift datasets. Our method can also be integrated with methods with label-dependent optimization to further push the performance boundary. Our code is available at https://github.com/ynanwu/MABN.

Distributional Reinforcement Learning (RL) estimates return distribution mainly by learning quantile values via minimizing the quantile Huber loss function, entailing a threshold parameter often selected heuristically or via hyperparameter search, which may not generalize well and can be suboptimal. This paper introduces a generalized quantile Huber loss function derived from Wasserstein distance (WD) calculation between Gaussian distributions, capturing noise in predicted (current) and target (Bellman-updated) quantile values. Compared to the classical quantile Huber loss, this innovative loss function enhances robustness against outliers. Notably, the classical Huber loss function can be seen as an approximation of our proposed loss, enabling parameter adjustment by approximating the amount of noise in the data during the learning process. Empirical tests on Atari games, a common application in distributional RL, and a recent hedging strategy using distributional RL, validate the effectiveness of our proposed loss function and its potential for parameter adjustments in distributional RL.

This paper proposes a learnable metric to measure the reconstruction quality of hyperspectral images obtained by computational hyperspectral imaging. Computational hyperspectral imaging aims to obtain low-cost hyperspectral images through consumer camera. While many hyperspectral reconstruction models have been developed for this purpose, conventional image and spectral quality metrics are insufficient to measure the scientific value of the reconstructed HSI cube. This paper proposes an adaptive quality fusion metric (AQF), adaptively aggregating the quality measures from point-wise, spatial-wise and spectral-wise aspects to assess the scientific value preserved by the reconstructed HSI. The proposed AQF metric uses weight parameters generated by a modified hypernetwork to determine the contribution for the three aspects given paired of groundtruth HSI and reconstructed HSI. Experimental results show its compatibility with existing metrics while accurately measuring the scientific information retained by the reconstructed HSI for hyperspectral applications.

Multi-label image classification presents a challenging task in many domains, including computer vision and medical imaging. Recent advancements have introduced graph-based and transformer-based methods to improve performance and capture label dependencies. However, these methods often include complex modules that entail heavy computation and lack interpretability. In this paper, we propose Probabilistic Multi-label Contrastive Learning (ProbMCL), a novel framework to address these challenges in multi-label image classification tasks. Our simple yet effective approach employs supervised contrastive learning, in which samples that share enough labels with an anchor image based on a decision threshold are introduced as a positive set. This structure captures label dependencies by pulling positive pair embeddings together and pushing away negative samples that fall below the threshold. We enhance representation learning by incorporating a mixture density network into contrastive learning and generating Gaussian mixture distributions to explore the epistemic uncertainty of the feature encoder. We validate the effectiveness of our framework through experimentation with datasets from the computer vision and medical imaging domains. Our method outperforms the existing state-of-the-art methods while achieving a low computational footprint on both datasets. Visualization analyses also demonstrate that ProbMCL-learned classifiers maintain a meaningful semantic topology.

Drawing inspiration from the primate brain’s intriguing evidence accumulation process, and guided by models from cognitive psychology and neuroscience, the paper introduces the NYCTALE framework, a neuro-inspired and evidence accumulation-based Transformer architecture. The proposed neuro-inspired NYCTALE offers a novel pathway in the domain of Personalized Medicine (PM) for lung cancer diagnosis. In nature, Nyctales are small owls known for their nocturnal behavior, hunting primarily during the darkness of night. The NYCTALE operates in a similarly vigilant manner, i.e., processing data in an evidence-based fashion and making predictions dynamically/adaptively. Distinct from conventional Computed Tomography (CT)-based Deep Learning (DL) models, the NYCTALE performs predictions only when sufficient amount of evidence is accumulated. In other words, instead of processing all or a pre-defined subset of CT slices, for each person, slices are provided one at a time. The NYCTALE framework then computes an evidence vector associated with contribution of each new CT image. A decision is made once the total accumulated evidence surpasses a specific threshold. Preliminary experimental analyses conducted using a challenging in-house dataset comprising 114 subjects. The results are noteworthy, suggesting that NYCTALE outperforms the benchmark accuracy even with approximately $60 \%$ less training data on this demanding and small dataset.

In this paper, we aim to adapt a model at test-time using a few unlabeled data to address distribution shifts. To tackle the challenges of extracting domain knowledge from a limited amount of data, it is crucial to utilize correlated information from pre-trained backbones and source domains. Previous studies fail to utilize recent foundation models with strong out-of-distribution generalization. Additionally, domain-centric designs are not flavored in their works. Furthermore, they employ the process of modelling source domains and the process of learning to adapt independently into disjoint training stages. In this work, we propose an approach on top of the pre-computed features of the foundation model. Specifically, we build a knowledge bank to learn the transferable knowledge from source domains. Conditioned on few-shot target data, we introduce a domain prompt generator to condense the knowledge bank into a domain-specific prompt. The domain prompt then directs the visual features towards a particular domain via a guidance module. Moreover, we propose a domain-aware contrastive loss and employ meta-learning to facilitate domain knowledge extraction. Extensive experiments are conducted to validate the domain knowledge extraction. The proposed method outperforms previous work on 5 large-scale benchmarks including WILDS and DomainNet.

We introduce the $\textbf{O}$ne-shot $\textbf{P}$runing $\textbf{T}$echnique for $\textbf{I}$nterchangeable $\textbf{N}$etworks ($\textbf{OPTIN}$) framework as a tool to increase the efficiency of pre-trained transformer architectures $\textit{without requiring re-training}$. Recent works have explored improving transformer efficiency, however often incur computationally expensive re-training procedures or depend on architecture-specific characteristics, thus impeding practical wide-scale adoption. To address these shortcomings, the OPTIN framework leverages intermediate feature distillation, capturing the long-range dependencies of model parameters (coined $\textit{trajectory}$), to produce state-of-the-art results on natural language, image classification, transfer learning, and semantic segmentation tasks $\textit{without re-training}$. Given a FLOP constraint, the OPTIN framework will compress the network while maintaining competitive accuracy performance and improved throughput. Particularly, we show a $\leq 2$% accuracy degradation from NLP baselines and a $0.5$% improvement from state-of-the-art methods on image classification at competitive FLOPs reductions. We further demonstrate the generalization of tasks and architecture with comparative performance using Mask2Former for semantic segmentation and cnn-style networks. OPTIN presents one of the first one-shot efficient frameworks for compressing transformer architectures that generalizes well across different class domains, in particular: natural language and image-related tasks, without $\textit{re-training}$.

The availability of large-scale multimodal datasets and advancements in diffusion models have significantly accelerated progress in 4D content generation. Most prior approaches rely on multiple images or video diffusion models, utilizing score distillation sampling for optimization or generating pseudo novel views for direct supervision. However, these methods are hindered by slow optimization speeds and multi-view inconsistency issues. Spatial and temporal consistency in 4D geometry has been extensively explored respectively in 3D-aware diffusion models and traditional monocular video diffusion models. Building on this foundation, we propose a strategy to migrate the temporal consistency in video diffusion models to the spatial-temporal consistency required for 4D generation. Specifically, we present a novel framework, \textbf{Diffusion4D}, for efficient and scalable 4D content generation. Leveraging a meticulously curated dynamic 3D dataset, we develop a 4D-aware video diffusion model capable of synthesizing orbital views of dynamic 3D assets. To control the dynamic strength of these assets, we introduce a 3D-to-4D motion magnitude metric as guidance. Additionally, we propose a novel motion magnitude reconstruction loss and 3D-aware classifier-free guidance to refine the learning and generation of motion dynamics. After obtaining orbital views of the 4D asset, we perform explicit 4D construction with Gaussian splatting in a coarse-to-fine manner. Extensive experiments demonstrate that our method surpasses prior state-of-the-art techniques in terms of generation efficiency and 4D geometry consistency across various prompt modalities.

Coded/uncoded content placement in Mobile Edge Caching (MEC) has evolved as an efficient solution to meet the significant growth of global mobile data traffic by boosting the content diversity in the storage of caching nodes. To meet the dynamic nature of the historical request pattern of multimedia contents, the main focus of recent researches has been shifted to develop data-driven and real-time caching schemes. In this regard and with the assumption that users’ preferences remain unchanged over a short horizon, the Top-K popular contents. These contents refer to the most requested content in the upcoming period. Most existing data-driven popularity prediction models, however, are not suitable for the coded/uncoded content placement frameworks. On the one hand, in coded/uncoded content placement, in addition to classifying contents into two groups, i.e., popular and non-popular, the probability of content request is required to identify which content should be stored partially/completely, where this information is not provided by existing data-driven popularity prediction models. On the other hand, the assumption that users’ preferences remain unchanged over a short horizon only works for content with a smooth request pattern. To tackle these challenges, we develop a Multiple-model (hybrid) Transformer-based Edge Caching (MTEC) framework with higher generalization ability, suitable for various types of content with different time-varying behavior, that can be adapted with coded/uncoded content placement frameworks. In this work, we consider Top-K content as the output of the 1st Stage of the proposed MTEC framework, which includes both popular and mediocre content. Simulation results corroborate the effectiveness of the proposed MTEC caching framework in comparison to its counterparts in terms of the cache-hit ratio, classification accuracy, and the transferred byte volume.

Mobile-edge caching (MEC) integrated with deep neural networks (DNNs) is an innovative technology with significant potential for the future generation of wireless networks, resulting in a considerable reduction in users’ latency. The mobile-edge caching (MEC) network’s effectiveness, however, heavily relies on its capacity to predict and dynamically update the storage of caching nodes with the most popular contents. To be effective, a DNN-based popularity prediction model needs to have the ability to understand the historical request patterns of content, including their temporal and spatial correlations. Existing state-of-the-art time-series DNN models capture the latter by simultaneously inputting the sequential request patterns of multiple contents to the network, considerably increasing the size of the input sample. This motivates us to address this challenge by proposing a DNN-based popularity prediction framework based on the idea of contrasting input samples against each other, designed for the unmanned aerial vehicle (UAV)-aided MEC networks. Referred to as the contrastive learning-based survival analysis (CLSA), the proposed architecture consists of a self-supervised contrastive learning (CL) model, where the temporal information of sequential requests is learned using a long short-term memory (LSTM) network as the encoder of the CL architecture. Followed by a survival analysis (SA) network, the output of the proposed CLSA architecture is probabilities for each content’s future popularity, which are then sorted in descending order to identify the Top-$K$popular contents. Based on the simulation results, the proposed CLSA architecture outperforms its counterparts across the classification accuracy and cache-hit ratio.

Reinforcement learning (RL) has garnered significant attention for developing decision-making agents that aim to maximize rewards, specified by an external supervisor, within fully observable environments. However, many real-world problems involve partial or noisy observations, where agents cannot access complete and accurate information about the environment. These problems are commonly formulated as partially observable Markov decision processes (POMDPs). Previous studies have tackled RL in POMDPs by either incorporating the memory of past actions and observations or by inferring the true state of the environment from observed data. Nevertheless, aggregating observations and actions over time becomes impractical in problems with large decision-making time horizons and high-dimensional spaces. Furthermore, inference-based RL approaches often require many environmental samples to perform well, as they focus solely on reward maximization and neglect uncertainty in the inferred state. Active inference (AIF) is a framework naturally formulated in POMDPs and directs agents to select actions by minimizing a function called expected free energy (EFE). This supplies reward-maximizing (or exploitative) behavior, as in RL, with information-seeking (or exploratory) behavior. Despite this exploratory behavior of AIF, its use is limited to problems with small time horizons and discrete spaces due to the computational challenges associated with EFE. In this article, we propose a unified principle that establishes a theoretical connection between AIF and RL, enabling seamless integration of these two approaches and overcoming their limitations in continuous space POMDP settings. We substantiate our findings with rigorous theoretical analysis, providing novel perspectives for using AIF in designing and implementing artificial agents. Experimental results demonstrate the superior learning capabilities of our method compared to other alternative RL approaches in solving partially observable tasks with continuous spaces. Notably, our approach harnesses information-seeking exploration, enabling it to effectively solve reward-free problems and rendering explicit task reward design by an external supervisor optional.

Differentiable Architecture Search (DARTS) has attracted considerable attention as a gradient-based neural architecture search method. Since the introduction of DARTS, there has been little work done on adapting the action space based on state-of-art architecture design principles for CNNs. In this work, we aim to address this gap by incrementally augmenting the DARTS search space with micro-design changes inspired by ConvNeXt and studying the trade-off between accuracy, evaluation layer count, and computational cost. We introduce the Pseudo-Inverted Bottleneck Conv (PIBConv) block intending to reduce the computational footprint of the inverted bottleneck block proposed in ConvNeXt. Our proposed architecture is much less sensitive to evaluation layer count and outperforms a DARTS network with similar size significantly, at layer counts as small as 2. Furthermore, with less layers, not only does it achieve higher accuracy with lower computational footprint (measured in GMACs) and parameter count, GradCAM comparisons show that our network can better detect distinctive features of target objects compared to DARTS. Code is available from https://github.com/mahdihosseini/PIBConv.

Mobile Edge Caching (MEC) is a revolutionary technology for the Sixth Generation (6G) of wireless networks with the promise to significantly reduce users’ latency via offering storage capacities at the edge of the network. The efficiency of the MEC network, however, critically depends on its ability to dynamically predict/update the storage of caching nodes with the top-K popular contents. Conventional statistical caching schemes are not robust to the time-variant nature of the underlying pattern of content requests, resulting in a surge of interest in using Deep Neural Networks (DNNs) for time-series popularity prediction in MEC networks. However, existing DNN models within the context of MEC fail to simultaneously capture both temporal correlations of historical request patterns and the dependencies between multiple contents. This necessitates an urgent quest to develop and design a new and innovative popularity prediction architecture to tackle this critical challenge. The paper addresses this gap by proposing a novel hybrid caching framework based on the attention mechanism. Referred to as the parallel Vision Transformers with Cross Attention (ViT-CAT) Fusion, the proposed architecture consists of two parallel ViT networks, one for collecting temporal correlation, and the other for capturing dependencies between different contents. Followed by a Cross Attention (CA) module as the Fusion Center (FC), the proposed ViT-CAT is capable of learning the mutual information between temporal and spatial correlations, as well, resulting in improving the classification accuracy, and decreasing the model’s complexity about 8 times. Based on the simulation results, the proposed ViT-CAT architecture outperforms its counterparts across the classification accuracy, complexity, and cache-hit ratio.

The paper proposes a novel hybrid discovery Radiomics framework that simultaneously integrates temporal and spatial features extracted from non-thin chest Computed Tomography (CT) slices to predict Lung Adenocarcinoma (LUAC) malignancy with minimum expert involvement. Lung cancer is the leading cause of mortality from cancer worldwide and has various histologic types, among which LUAC has recently been the most prevalent. LUACs are classified as pre-invasive, minimally invasive, and invasive adenocarcinomas. Timely and accurate knowledge of the lung nodules malignancy leads to a proper treatment plan and reduces the risk of unnecessary or late surgeries. Currently, chest CT scan is the primary imaging modality to assess and predict the invasiveness of LUACs. However, the radiologists’ analysis based on CT images is subjective and suffers from a low accuracy compared to the ground truth pathological reviews provided after surgical resections. The proposed hybrid framework, referred to as the CAET-SWin, consists of two parallel paths: (i) The Convolutional Auto-Encoder (CAE) Transformer path that extracts and captures informative features related to inter-slice relations via a modified Transformer architecture, and; (ii) The Shifted Window (SWin) Transformer path, which is a hierarchical vision transformer that extracts nodules’ related spatial features from a volumetric CT scan. Extracted temporal (from the CAET path) and spatial (from the SWin path) are then fused through a fusion path to classify LUACs. Experimental results on our in-house dataset of 114 pathologically proven SubSolid Nodules (SSNs) demonstrate that the CAET-SWin significantly improves reliability of the invasiveness prediction task while achieving an accuracy of 82.65%, sensitivity of 83.66%, and specificity of 81.66% using 10-fold cross-validation.

Exploration is a significant challenge in practical reinforcement learning (RL), and uncertainty-aware exploration that incorporates the quantification of epistemic and aleatory uncertainty has been recognized as an effective exploration strategy. However, capturing the combined effect of aleatory and epistemic uncertainty for decision-making is difficult. Existing works estimate aleatory and epistemic uncertainty separately and consider the composite uncertainty as an additive combination of the two. Nevertheless, the additive formulation leads to excessive risk-taking behavior, causing instability. In this paper, we propose an algorithm that clarifies the theoretical connection between aleatory and epistemic uncertainty, unifies aleatory and epistemic uncertainty estimation, and quantifies the combined effect of both uncertainties for a risk-sensitive exploration. Our method builds on a novel extension of distributional RL that estimates a parameterized return distribution whose parameters are random variables encoding epistemic uncertainty. Experimental results on tasks with exploration and risk challenges show that our method outperforms alternative approaches.

This paper presents a new distance metric to compare two continuous probability density functions. The main advantage of this metric is that, unlike other statistical measurements, it can provide an analytic, closed-form expression for a mixture of Gaussian distributions while satisfying all metric properties. These characteristics enable fast, stable, and efficient calculations, which are highly desirable in real-world signal processing applications. The application in mind is Gaussian Mixture Reduction (GMR), which is widely used in density estimation, recursive tracking, and belief propagation. To address this problem, we developed a novel algorithm dubbed the Optimization-based Greedy GMR (OGGMR), which employs our metric as a criterion to approximate a high-order Gaussian mixture with a lower order. Experimental results show that the OGGMR algorithm is significantly faster and more efficient than state-of-the-art GMR algorithms while retaining the geometric shape of the original mixture.

Researchers have long tried to minimize training costs in deep learning while maintaining strong generalization across diverse datasets. Emerging research on dataset distillation aims to reduce training costs by creating a small synthetic set that contains the information of a larger real dataset and ultimately achieves test accuracy equivalent to a model trained on the whole dataset. Unfortunately, the synthetic data generated by previous methods are not guaranteed to distribute and discriminate as well as the original training data, and they incur significant computational costs. Despite promising results, there still exists a significant performance gap between models trained on condensed synthetic sets and those trained on the whole dataset. In this paper, we address these challenges using efficient Dataset Distillation with Attention Matching (DataDAM), achieving state-of-the-art performance while reducing training costs. Specifically, we learn synthetic images by matching the spatial attention maps of real and synthetic data generated by different layers within a family of randomly initialized neural networks. Our method outperforms the prior methods on several datasets, including CIFAR10/100, TinyImageNet, ImageNet-1K, and subsets of ImageNet-1K across most of the settings, and achieves improvements of up to 6.5% and 4.1% on CIFAR100 and ImageNet-1K, respectively. We also show that our high-quality distilled images have practical benefits for downstream applications, such as continual learning and neural architecture search.

We introduce Hyper-Skin, a hyperspectral dataset covering wide range of wavelengths from visible (VIS) spectrum (400nm - 700nm) to near-infrared (NIR) spectrum (700nm - 1000nm), uniquely designed to facilitate research on facial skin-spectra reconstruction.By reconstructing skin spectra from RGB images, our dataset enables the study of hyperspectral skin analysis, such as melanin and hemoglobin concentrations, directly on the consumer device. Overcoming limitations of existing datasets, Hyper-Skin consists of diverse facial skin data collected with a pushbroom hyperspectral camera. With 330 hyperspectral cubes from 51 subjects, the dataset covers the facial skin from different angles and facial poses.Each hyperspectral cube has dimensions of 1024$\times$1024$\times$448, resulting in millions of spectra vectors per image. The dataset, carefully curated in adherence to ethical guidelines, includes paired hyperspectral images and synthetic RGB images generated using real camera responses. We demonstrate the efficacy of our dataset by showcasing skin spectra reconstruction using state-of-the-art models on 31 bands of hyperspectral data resampled in the VIS and NIR spectrum. This Hyper-Skin dataset would be a valuable resource to NeurIPS community, encouraging the development of novel algorithms for skin spectral reconstruction while fostering interdisciplinary collaboration in hyperspectral skin analysis related to cosmetology and skin's well-being. Instructions to request the data and the related benchmarking codes are publicly available at: https://github.com/hyperspectral-skin/Hyper-Skin-2023.

This article deals with the problem of statistical and system heterogeneity in a cross-silo federated learning (FL) framework where there exist a limited number of Consumer Internet of Things (CIoT) devices in a smart building. We propose a novel graph signal processing (GSP)-inspired aggregation rule based on graph filtering dubbed “G-Fedfilt.” The proposed aggregator enables a structured flow of information based on the graph’s topology. This behavior allows capturing the interconnection of CIoT devices and training domain-specific models. The embedded graph filter is equipped with a tunable parameter which enables a continuous tradeoff between domain-agnostic and domain-specific FL. In the case of domain-agnostic, it forces G-Fedfilt to act similar to the conventional federated averaging (FedAvg) aggregation rule. The proposed G-Fedfilt also enables an intrinsic smooth clustering based on the graph connectivity without explicitly specified which further boosts the personalization of the models in the framework. In addition, the proposed scheme enjoys a communication-efficient time scheduling to alleviate the system heterogeneity. This is accomplished by adaptively adjusting the amount of training data samples and sparsity of the models’ gradients to reduce communication desynchronization and latency. Simulation results show that the proposed G-Fedfilt achieves up to 3.99% better classification accuracy than the conventional FedAvg when concerning model personalization on the statistically heterogeneous local data sets, while it is capable of yielding up to 2.41% higher accuracy than FedAvg in the case of testing the generalization of the models. Furthermore, the proposed communication optimization scheme can boost the framework’s efficiency by reducing the computation, communication desynchronization, and latency up to 70.21%, 99.65%, and 44.61%, respectively, at the cost of 0.36% accuracy and under the system heterogeneity.

Over the years, the popularity and usage of wearable Internet of Things (IoT) devices in several healthcare services are increased. Among the services that benefit from the usage of such devices is predictive analysis, which can improve early diagnosis in e-health. However, due to the limitations of wearable IoT devices, challenges in data privacy, service integrity, and network structure adaptability arose. To address these concerns, we propose a platform using federated learning and private blockchain technology within a fog-IoT network. These technologies have privacy-preserving features securing data within the network. We utilized the fog-IoT network’s distributive structure to create an adaptive network for wearable IoT devices. We designed a testbed to examine the proposed platform’s ability to preserve the integrity of a classifier. According to experimental results, the introduced implementation can effectively preserve a patient’s privacy and a predictive service’s integrity. We further investigated the contributions of other technologies to the security and adaptability of the IoT network. Overall, we proved the feasibility of our platform in addressing significant security and privacy challenges of wearable IoT devices in predictive healthcare through analysis, simulation, and experimentation.

Stress is an inevitable part of modern life. While stress can negatively impact a person's life and health, positive and under-controlled stress can also enable people to generate creative solutions to problems encountered in their daily lives. Although it is hard to eliminate stress, we can learn to monitor and control its physical and psychological effects. It is essential to provide feasible and immediate solutions for more mental health counselling and support programs to help people relieve stress and improve their mental health. Popular wearable devices, such as smartwatches with several sensing capabilities, including physiological signal monitoring, can alleviate the problem. This work investigates the feasibility of using wrist-based electrodermal activity (EDA) signals collected from wearable devices to predict people's stress status and identify possible factors impacting stress classification accuracy. We use data collected from wrist-worn devices to examine the binary classification discriminating stress from non-stress. For efficient classification, five machine learning-based classifiers were examined. We explore the classification performance on four available EDA databases under different feature selections. According to the results, Support Vector Machine (SVM) outperforms the other machine learning approaches with an accuracy of 92.9 for stress prediction. Additionally, when the subject classification included gender information, the performance analysis showed significant differences between males and females. We further examine a multimodal approach for stress classifications. The results indicate that wearable devices with EDA sensors have a great potential to provide helpful insight for improved mental health monitoring.

This paper presents a nonlinear location estimation to infer the position of a user holding a smartphone. We consider a large location with$M$number of grid points, each grid point is labeled with a unique fingerprint consisting of the received signal strength (RSS) values measured from$N$number of Bluetooth Low Energy (BLE) beacons. Given the fingerprint observed by the smartphone, the user’s current location can be estimated by finding the top-k similar fingerprints from the list of fingerprints registered in the database. Besides the environmental factors, the dynamicity in holding the smartphone is another source to the variation in fingerprint measurements, yet there are not many studies addressing the fingerprint variability due to dynamic smartphone positions held by human hands during online detection. To this end, we propose a nonlinear location estimation using the kernel method. Specifically, our proposed method comprises of two steps: 1) a beacon selection strategy to select a subset of beacons that is insensitive to the subtle change of holding positions, and 2) a kernel method to compute the similarity between this subset of observed signals and all the fingerprints registered in the database. The experimental results based on large-scale data collected in a complex building indicate a substantial performance gain of our proposed approach in comparison to state-of-the-art methods. The dataset consisting of the signal information collected from the beacons is available online.

This paper investigates an Unmanned Aerial Vehicle (UAV)-enabled network consisting of smart mobile devices and multiple UAVs as aerial base stations in a Multiple-Input Multiple-Output (MIMO) architecture. Mobile devices are partitioned into several clusters and offload their tasks to the UAV servers via the Non-Orthogonal Multiple Access (NOMA) protocol. The main goal of the paper is to jointly maximize the number of served terrestrial users and their scheduling. Moreover, the number of UAV servers and their 3D placement are optimized. To this end, we formulate an optimization problem subject to some Quality of Service (QoS) constraints. The resulting problem is non-convex and intractable to solve. Therefore, we break the problem into two subproblems. We propose an efficient algorithm based on machine learning to solve the first subproblem, i.e., optimizing the number of UAVs and their 3D placements, and the user association. Different from existing literature, our proposed algorithm can achieve low computational complexity and fast convergence. The second subproblem, the user scheduling, is non-convex too. We utilize the$\ell _p$-norm concept to find a convex upper bound for the subproblem and optimize the user scheduling by applying the Successive Convex Approximation (SCA) algorithm. The aforementioned process is performed iteratively until the overall algorithm converges and a near-optimal solution is achieved for the optimization problem. Moreover, the computational complexity of the proposed scheme is analyzed. Finally, we evaluate the performance of our proposed algorithm via the simulation results. Regarding fast convergence and low computational complexity of the proposed algorithm, its superior performance is confirmed through numerical results.

Explaining the generalization characteristics of deep learning is an emerging topic in advanced machine learning. There are several unanswered questions about how learning under stochastic optimization really works and why certain strategies are better than others. In this paper, we address the following question: can we probe intermediate layers of a deep neural network to identify and quantify the learning quality of each layer? With this question in mind, we propose new explainability metrics that measure the redundant information in a network's layers using a low-rank factorization framework and quantify a complexity measure that is highly correlated with the generalization performance of a given optimizer, network, and dataset. We subsequently exploit these metrics to augment the Stochastic Gradient Descent (SGD) optimizer by adaptively adjusting the learning rate in each layer to improve in generalization performance. Our augmented SGD - dubbed RMSGD - introduces minimal computational overhead compared to SOTA methods and outperforms them by exhibiting strong generalization characteristics across application, architecture, and dataset.

This paper presents a human activity recognition (HAR) system with wearable devices. While various approaches have been suggested for HAR, most of them focus on either 1) the inertial sensors to capture the physical movement or 2) subject-dependent evaluations that are less practical to real world cases. To this end, our work integrates sensing in-puts from physiological sensors to compensate the limitation of inertial sensors in capturing the human activities with less physical movements. Physiological sensors can capture physiological responses reflecting human behaviors in executing daily activities. To simulate a realistic application, three different evaluation scenarios are considered, namely All-access, Cross-subject and Cross-activity. Lastly, we propose a Hierarchical Deep Learning (HDL) model, which improves the accuracy and stability of HAR, compared to conventional models. Our proposed HDL with fusion of inertial and physiological sensing inputs achieves 97.16%, 92.23%, 90.18% average accuracy in All-access, Cross-subject, Cross-activity scenarios, which confirms the effectiveness of our approach.

The lack of well-annotated datasets in computational pathology (CPath) obstructs the application of deep learning techniques for classifying medical images. Many CPath workflows involve transferring learned knowledge between various image domains through transfer learning. Currently, most transfer learning research follows a model-centric approach, tuning network parameters to improve transfer results over few datasets. In this paper, we take a data-centric approach to the transfer learning problem and examine the existence of generalizable knowledge between histopathological datasets. First, we create a standardization workflow for aggregating existing histopathological data. We then measure inter-domain knowledge by training ResNet18 models across multiple histopathological datasets, and cross-transferring between them to determine the quantity and quality of innate shared knowledge. Additionally, we use weight distillation to share knowledge between models without additional training. We find that hard to learn, multi-class datasets benefit most from pretraining, and a two stage learning framework incorporating a large source domain such as ImageNet allows for better utilization of smaller datasets. Furthermore, we find that weight distillation enables models trained on purely histopathological features to outperform models using external natural image data.

As a consequence of the COVID-19 pandemic, the demand for telecommunication for remote learning/working and telemedicine has significantly increased. Mobile Edge Caching (MEC) in the 6G networks has been evolved as an efficient solution to meet the phenomenal growth of the global mobile data traffic by bringing multimedia content closer to the users. Although massive connectivity enabled by MEC networks will significantly increase the quality of communications, there are several key challenges ahead. The limited storage of edge nodes, the large size of multimedia content, and the time-variant users’ preferences make it critical to efficiently and dynamically predict the popularity of content to store the most upcoming requested ones before being requested. Recent advancements in Deep Neural Networks (DNNs) have drawn much research attention to predict the content popularity in proactive caching schemes. Existing DNN models in this context, however, suffer from long-term dependencies, computational complexity, and unsuitability for parallel computing. To tackle these challenges, we propose an edge caching framework incorporated with the attention-based Vision Transformer (ViT) neural network, referred to as the Transformer-based Edge (TEDGE) caching, which to the best of our knowledge, is being studied for the first time. Moreover, the TEDGE caching framework requires no data pre-processing and additional contextual information. Simulation results corroborate the effectiveness of the proposed TEDGE caching framework in comparison to its counterparts.

The recent pandemic has brought tremendous changes to everyone’s life, causing stress about losing loved ones, losing jobs, and having changes in sleep or eating habits. This study investigates the feasibility of utilizing Electrodermal Activity (EDA) collected from wearable devices to detect people’s stress. EDA can quantify the changes in sympathetic dynamics by measuring sweat produced by our sweat glands. Currently, the adoption of EDA sensors to commercially off-the-shelf smart-watches is still in the infancy stage, and only a few brands have the EDA sensors implemented into their smartwatch. To facilitate our feasibility study, we need the datasets that contain the EDA signals collected from wearable devices. This paper uses two publicly available datasets containing the EDA signals collected from research-grade wearable devices. We cast the stress detection problem as a binary classification problem and trained the classifiers with three popular machine learning methods: K-Nearest Neighbor, Logistic Regression, and Random Forests. According to experimental results, Random Forests achieves an accuracy of 85.7% to classify stress from non-stress status. The results verified that wearable devices with EDA sensors have the potential to predict stress status.

Multi-Agent Reinforcement Learning (MARL) has demonstrated significant suc2 cess by virtue of collaboration across agents. Recent work, on the other hand, introduces surprise which quantifies the degree of change in an agent’s environ4 ment. Surprise-based learning has received significant attention in the case of single-agent entropic settings but remains an open problem for fast-paced dynamics in multi-agent scenarios. A potential alternative to address surprise may be realized through the lens of free-energy minimization. We explore surprise minimization in multi-agent learning by utilizing the free energy across all agents in a multi-agent system. A temporal Energy-Based Model (EBM) represents an estimate of surprise which is minimized over the joint agent distribution. Our formulation of the EBM is theoretically akin to the minimum conjugate entropy objective and highlights suitable convergence towards minimum surprising states. We further validate our theoretical claims in an empirical study of multi-agent tasks demanding collabora14 tion in the presence of fast-paced dynamics. Our implementation and agent videos are available at the Project Webpage.

The number of embedded devices that connect to a wireless network has been growing for the past decade. This interaction creates a network of Internet-of-Things (IoT) devices where data travel continuously. With the increase of devices and the need for the network to extend via fog computing, we have fog-based IoT networks. However, with more endpoints introduced to it, the network becomes open to malicious attackers. This work attempts to protect fog-based IoT networks by creating a platform that secures the endpoints through public-key encryption. The servers are allowed to mask the data packets shared within the network. To be able to track all of the encryption processes, we incorporated the use of permissioned blockchains. This technology completes the security layer by providing an immutable and automated data structure to function as a hyper ledger for the network. Each data transaction incorporates a handshake mechanism with the use of a public-key pair. This design guarantees that only devices that have proper access through the keys can use the network. Hence, management is made convenient and secure. The implementation of this platform is through a wireless server–client architecture to simulate the data transactions between devices. The conducted qualitative tests provide an in-depth feasibility investigation on the network’s levels of security. The results show the validity of the design as a means of fortifying the network against endpoint attacks.

This article considers an ultradense heterogeneous network (UDHN) consisting of cache-enabled unmanned aerial vehicles (UAVs) and Internet of Things mobile devices (IMDs) receiving their requested contents via the power domain nonorthogonal multiple access (PD-NOMA) protocol. Employing the fast global${K}$-means (FGKMs) algorithm, IMDs are partitioned into several clusters connecting to either other IMDs or UAV belonging to the same cluster in the presence of untrusted users, known as nonlegitimate eavesdroppers. It is assumed that users located in the cluster edge area communicate with multiple UAVs to obtain their requested contents. The main goal for such a network is to jointly optimize the number of UAVs, their 3-D placements, and the cache placement probability of contents stored in UAVs and IMDs by maximizing the secure cache throughput. Toward this goal, we prove that the objective function is nonconcave. Therefore, we decompose the optimization problem into multiple subproblems. We first employ the FGKM algorithm to optimally determine the number of employed UAVs and their horizontal placements. Then, the UAVs’ altitudes are optimized by employing the interior-point method (IPM), while the convex approximation for the objective function and its constraints are substituted. Then, we optimize the secure cache throughput of IMDs by proposing a caching placement strategy for contents stored in UAVs and IMDs via Device to Device (D2D) and UAV to Device (U2D) links, given illegal eavesdroppers’ presence. Then, we propose an iterative algorithm to achieve the near-optimal solution for the cache throughput of IMDs. Different from existing works, the closed-form expressions for the achievable secrecy rate and the successful probability of D2D communications and U2D transmissions, as well as the secure cache throughput are derived. Finally, simulation results are presented to validate the proposed caching placement strategy. It is shown that the proposed scheme outperforms the conventional most popular caching (MPC) strategy substantially.

Recently, as a consequence of the COVID-19 pandemic, dependence on telecommunication for remote learning/working and telemedicine has significantly increased. In this context, preserving high Quality of Service (QoS) and maintaining low-latency communication are of paramount importance. In cellular networks, the incorporation of unmanned aerial vehicles (UAVs) can result in enhanced connectivity for outdoor users due to the high probability of establishing Line of Sight (LoS) links. The UAV’s limited battery life and its signal attenuation in indoor areas, however, make it inefficient to manage users’ requests in indoor environments. Referred to as the cluster-centric and coded UAV-aided femtocaching (CCUF) framework, the network’s coverage in both indoor and outdoor environments increases by considering a two-phase clustering framework for Femto access points (FAPs)’ formation and UAVs’ deployment. Our first objective is to increase the content diversity. In this context, we propose a coded content placement in a cluster-centric cellular network, which is integrated with the coordinated multipoint (CoMP) approach to mitigate the intercell interference in edge areas. Then, we compute, experimentally, the number of coded contents to be stored in each caching node to increase the cache-hit-ratio, signal-to-interference-plus-noise ratio (SINR), and cache diversity and decrease the users’ access delay and cache redundancy for different content popularity profiles. Capitalizing on clustering, our second objective is to assign the best caching node to indoor/outdoor users for managing their requests. In this regard, we define the movement speed of ground users as the decision metric of the transmission scheme for serving outdoor users’ requests to avoid frequent handovers between FAPs and increase the battery life of UAVs. Simulation results illustrate that the proposed CCUF implementation increases the cache-hit-ratio, SINR, and cache diversity and decrease the users’ access delay, cache redundancy, and UAVs’ energy consumption.

Vehicular networks are known for their dynamic topology, high mobility, and frequent disconnections. Unmanned aerial vehicles (UAVs) have been recently used as instant communication relays to bridge the communication gaps between terrestrial vehicles to improve connectivity in vehicular networks and overcome the aforementioned problems. Despite the existing work in the literature where each vehicle connects directly to UAVs, this work studies how clustering and different densities of vehicles affect delay and energy efficiency in a UAV-based vehicular network integrated with 5G technology. Consequently, this work addresses the problem of UAV enabling vehicular ad-hoc networks (VANETs) in a highway scenario, where UAVs serve as an effective complement to forward data packets between vehicles in the absence of sufficient fixed infrastructures in an emergency situation. The main objective of this work is to minimize the delay while maximizing the energy efficiency by minimizing the power consumption and maximizing the total data rate under realistic conditions, while nonorthogonal multiple access (NOMA) is also adopted as an alternative answer for the effective utilization of limited bandwidth. The free-flowing traffic follows a Poisson stochastic process where each vehicle is assigned a random speed selected from a truncated Gaussian distribution. To this end, a novel modification of fast global$K$-means is adopted to partition vehicles, allowing communication between clusters by vehicle-to-vehicle links, while data packets between clusters are relayed through UAVs. By computing the convex approximation of the objective function and the constraints, the original problem with the mixed-integer, nonconvex, and nonlinear form is solved by the proposed iterative inner penalty function algorithm. Finally, extensive simulations are conducted to validate the superiority of the proposed method in terms of various metrics. The results indicate that the relaying task in the proposed UAV-assisted VANET based on 5G technology is perfectly suited to enhance the network connectivity.

This article considers a wireless network consisting of unmanned aerial vehicles (UAVs), deployed as aerial base stations, and a large number of terrestrial users randomly distributed in a dense urban area. The main objective of this work is to maximize the downlink rate of users along with clustering of users and 2-D initial placement of UAVs, which effectively minimizes the clustering error. To achieve this goal, we estimate the next users’ locations with deep echo-state network (ESN) to find the movement pattern of users with high accuracy. Then, we propose the single- and multiagent actor–critic (AC) algorithms for UAVs’ initial deployment and trajectory design, where the multiagent scheme employs an efficient bandwidth allocation. Simulation results supported by a real data set of the terrestrial users’ coordinates indicate that, when the deep ESN algorithm is used, the accuracy is 93.75% for longitude and 88.36% for latitude compared to the simple ESN performance. Moreover, the use of single- and multiagent AC algorithms display better performance in terms of downlink rate and convergence speed than value-based algorithms such as deep$Q$-network schemes.

This article considers the problem of an unmanned aerial vehicle (UAV)-enabled cloud network under partial computation offloading scenario, where multiple UAV-mounted aerial base stations are employed to serve a group of remote Internet-of-Things ground-based smart devices (ISDs). The main objective of this work is to maximize energy efficiency by minimizing the number of needed drones while minimizing the cost associated with serving the ISDs under some realistic quality of service constraints. To that end, we aim to jointly optimize the 3-D UAV placements, transmit power, and cloud resources. This represents a challenging, nonconvex, and NP-hard optimization problem. In this work, we decompose the optimization problem into three separate subproblems, namely, 2-D UAV positioning, UAV altitude optimization, and UAV-cloud resource association. These subproblems are solved using a modified global$K$-means, successive convex approximation, and successive linear programming techniques. A comprehensive simulation study and comparative evaluation against the state-of-the-art (SOTA) algorithms are conducted to demonstrate the utility of the proposed approach and its benefits in applications of interest.

Artificial intelligence (AI) and 5G system have been two hot technical areas that are changing the world. On the deep convergence of computing and communication, networking systems of AI (NSAI) is presenting a paradigm shift, where distributed AI becomes immersive in all elements of the network, i.e., cloud, edge, and terminal devices, which make AI virtually operating as a networking system. On the other hand, by the evolution of the communication systems, a network is becoming a service-specific system interweaved with AI, i.e., the network operates as an AI system, enabling real-time smart services. With the developing technology trends of “AI as a network and network as an AI,” the ecosystem of NSAI can be presenting the next-generation waves of both AI systems and B5G-6G communication networks. In this article, we mainly aim to provide a comprehensive survey on the system architecture, key technologies, application scenarios, challenges, and opportunities of NSAI, which can shed light on the future developments of both telecommunications and AI computing. The contributions of this article also include: 1) providing a unified framework for the deep convergence of computing and communications, where the network and application/service can be jointly optimized as a single integrated system and 2) suggesting the roadmap and open research problems in realizing the online-evolutive integration of cyberspace, physical world, and human society, toward the ubiquitous brain networks (UBNs), which are requiring the joint efforts from both research communities of computing and communication.

As an emerging field in Machine Learning, Explainable AI (XAI) has been offering remarkable performance in interpreting the decisions made by Convolutional Neural Networks (CNNs). To achieve visual explanations for CNNs, methods based on class activation mapping and randomized input sampling have gained great popularity. However, the attribution methods based on these techniques provide lower-resolution and blurry explanation maps that limit their explanation power. To circumvent this issue, visualization based on various layers is sought. In this work, we collect visualization maps from multiple layers of the model based on an attribution-based input sampling technique and aggregate them to reach a fine-grained and complete explanation. We also propose a layer selection strategy that applies to the whole family of CNN-based models, based on which our extraction framework is applied to visualize the last layers of each convolutional block of the model. Moreover, we perform an empirical analysis of the efficacy of derived lower-level information to enhance the represented attributions. Comprehensive experiments conducted on shallow and deep models trained on natural and industrial datasets, using both ground-truth and model-truth based evaluation metrics validate our proposed algorithm by meeting or outperforming the state-of-the-art methods in terms of explanation ability and visual quality, demonstrating that our method shows stability regardless of the size of objects or instances to be explained.

Computer Aided Diagnosis (CAD) systems are increasingly utilizing image analysis and Deep Learning (DL) techniques, due to their high accuracy in several medical imaging fields, including the detection of Acute Lymphoblastic (or Lymphocytic) Leukemia (ALL) from peripheral blood samples. However, no method in the literature has specifically analyzed the focus quality of ALL images or proposed a technique for sharpening the samples in an adaptive way for the purpose of classification. To address this issue, in this paper we propose the first machine learning-based approach able to enhance blood sample images by an adaptive unsharpening method. The method uses image processing techniques and DL to normalize the radius of the cell, estimate the focus quality, adaptively improve the sharpness of the images, and then perform the classification. We evaluated the methodology on a public database of ALL images, considering several state-of-the-art CNNs to perform the classification, with results showing the validity of the proposed approach. For a complete reproducibility of the work, the source code is available at: http://iebil.di.unimi.it/cnnALL/index.htm.

Bluetooth Low Energy (BLE) is one of the key technologies empowering the Internet of Things (IoT) for indoor positioning. In this regard, Angle of Arrival (AoA) localization is one of the most reliable techniques because of its low estimation error. BLE-based AoA localization, however, is in its infancy as only recently direction-finding feature is introduced to the BLE specification. Furthermore, AoA-based approaches are prone to noise, multi-path, and path-loss effects. The paper proposes an efficient Convolutional Neural Network (CNN)-based indoor localization framework to tackle these issues specific to BLE-based settings. We consider indoor environments without presence of Line of Sight (LoS) links affected by Additive White Gaussian Noise (AWGN) with different Signal to Noise Ratios (SNRs) and Rayleigh fading channel. Moreover, by assuming a 3-D indoor environment, the destructive effect of the elevation angle of the incident signal is considered on the position estimation. The effectiveness of the proposed CNN-AoA framework is evaluated via an experimental testbed, where In-phase/Quadrature (I/Q) samples, modulated by Gaussian Frequency Shift Keying (GFSK), are collected by four BLE beacons. Simulation results corroborate effectiveness of the proposed CNN-based AoA technique to track mobile agents with high accuracy in the presence of noise and Rayleigh fading channel.

The global outbreak of the novel corona virus (COVID-19) disease has drastically impacted the world and led to one of the most challenging crisis across the globe since World War II. The early diagnosis and isolation of COVID-19 positive cases are considered as crucial steps towards preventing the spread of the disease and flattening the epidemic curve. Chest Computed Tomography (CT) scan is a highly sensitive, rapid, and accurate diagnostic technique that can complement Reverse Transcription Polymerase Chain Reaction (RT-PCR) test. Recently, deep learning-based models, mostly based on Convolutional Neural Networks (CNN), have shown promising diagnostic results. CNNs, however, are incapable of capturing spatial relations between image instances and require large datasets. Capsule Networks, on the other hand, can capture spatial relations, require smaller datasets, and have considerably fewer parameters. In this paper, a Capsule network framework, referred to as the "CT-CAPS", is presented to automatically extract distinctive features of chest CT scans. These features, which are extracted from the layer before the final capsule layer, are then leveraged to differentiate COVID-19 from Non-COVID cases. The experiments on our in-house dataset of 307 patients show the state-of-the-art performance with the accuracy of 90.8%, sensitivity of 94.5%, and specificity of 86.0%.

The paper is motivated by the importance of the Smart Cities (SC) concept for future management of global urbanization and energy consumption. Multi-agent Reinforcement Learning (RL) is an efficient solution to utilize large amount of sensory data provided by the Internet of Things (IoT) infrastructure of the SCs for city-wide decision making and managing demand response. Conventional ModelFree (MF) and Model-Based (MB) RL algorithms, however, use a fixed reward model to learn the value function rendering their application challenging for ever changing SC environments. Successor Representations (SR)-based techniques are attractive alternatives that address this issue by learning the expected discounted future state occupancy, referred to as the SR, and the immediate reward of each state. SR-based approaches are, however, mainly developed for single agent scenarios and have not yet been extended to multi-agent settings. The paper addresses this gap and proposes the Multi-Agent Adaptive Kalman Filtering-based Successor Representation (MAKF-SR) framework. The proposed framework can adapt quickly to the changes in a multi-agent environment faster than the MF methods and with a lower computational cost compared to MB algorithms. The proposed MAKF-SR is evaluated through a comprehensive set of experiments illustrating superior performance compared to its counterparts.

Visualizing the features captured by Convolutional Neural Networks (CNNs) is one of the conventional approaches to interpret the predictions made by these models in numerous image recognition applications. Grad-CAM is a popular solution that provides such a visualization by combining the activation maps obtained from the model. However, the average gradient-based terms deployed in this method underestimates the contribution of the representations discovered by the model to its predictions. Addressing this problem, we introduce a solution to tackle this issue by computing the path integral of the gradient-based terms in Grad-CAM. We conduct a thorough analysis to demonstrate the improvement achieved by our method in measuring the importance of the extracted representations for the CNN’s predictions, which yields to our method’s administration in object localization and model interpretation.

Explainable AI (XAI) is an active research area to interpret a neural network’s decision by ensuring transparency and trust in the task-specified learned models. Recently, perturbation-based model analysis has shown better interpretation, but backpropagation techniques are still prevailing because of their computational efficiency. In this work, we combine both approaches as a hybrid visual explanation algorithm and propose an efficient interpretation method for convolutional neural networks. Our method adaptively selects the most critical features that mainly contribute towards a prediction to probe the model by finding the activated features. Experimental results show that the proposed method can reduce the execution time up to 30% while enhancing competitive interpretability without compromising the quality of explanation generated.

The unprecedented COVID-19 pandemic has been remarkably impacting the world and influencing a broad aspect of people’s lives since its first emergence in late 2019. The highly contagious nature of the COVID-19 has raised the necessity of developing deep learning-based diagnostic tools to identify the infected cases in the early stages. Recently, we proposed a fully-automated framework based on Capsule Networks, referred to as the CT-CAPS, to distinguish COVID-19 infection from normal and Community Acquired Pneumonia (CAP) cases using chest Computed Tomography (CT) scans. Although CT scans can provide a comprehensive illustration of the lung abnormalities, COVID-19 lung manifestations highly overlap with the CAP findings making their identification challenging even for experienced radiologists. Here, the CT-CAPS is augmented with a wide range of clinical/demographic data, including patients’ gender, age, weight and symptoms. More specifically, we propose a hybrid deep learning model that utilizes both clinical/demographic data and CT scans to classify COVID-19 and non-COVID cases using a Random Forest Classifier. The proposed hybrid model specifies the most important predictive factors increasing the explainability of the model. The experimental results show that the proposed hybrid model improves the CT-CAPS performance, achieving accuracy of 90.8%, sensitivity of 94.5% and specificity of 86.0%.

Understanding the generalization behaviour of deep neural networks is a topic of recent interest that has driven the production of many studies, notably the development and evaluation of generalization “explainability” measures that quantity model generalization ability. Generalization measures have also proven useful in the development of powerful layer-wise model tuning and optimization algorithms, though these algorithms require specific kinds of generalization measures which can probe individual layers. The purpose of this paper is to explore the neglected subtopic of probeable generalization measures; to establish firm ground for further investigations, and to inspire and guide the development of novel model tuning and optimization algorithms. We evaluate and compare measures, demonstrating effectiveness and robustness across model variations, dataset complexities, training hyperparameters, and training stages. We also introduce a new dataset of trained models and performance metrics, GenProb, for testing generalization measures, model tuning algorithms and optimization algorithms.

Neural Architecture Search (NAS) has been pivotal in finding optimal network configurations for Convolution Neural Networks (CNNs). While many methods explore NAS from a global search-space perspective, the employed optimization schemes typically require heavy computational resources. This work introduces a method that is efficient in computationally constrained environments by examining the micro-search space of channel size. In tackling channel-size optimization, we design an automated algorithm to extract the dependencies within different connected layers of the network. In addition, we introduce the idea of knowledge distillation, which enables preservation of trained weights, admist trials where the channel sizes are changing. Further, since the standard performance indicators (accuracy, loss) fail to capture the performance of individual network components (providing an overall network evaluation), we introduce a novel metric that highly correlates with test accuracy and enables analysis of individual network layers. Combining dependency extraction, metrics, and knowledge distillation, we introduce an efficient searching algorithm, with simulated annealing inspired stochasticity, and demonstrate its effectiveness in finding optimal architectures that outperform baselines by a large margin.

Given the data-hungry nature of Wireless Multimedia Sensor Networks (WMSNs) due to the need for near real-time processing of a large number of multimedia data, it is of significant practical importance to design/develop energy-efficient routing protocols to extend the WMSN’s collective lifetime. In this regard and to jointly utilize potential benefits that can be achieved by coupling clustering and image compression, the paper proposes a novel routing methodology referred to as the Energy Efficient Cluster-based Image Transmission (EECIT) scheme. In the proposed EECIT scheme, the multimedia-based sensor network is divided into different clusters depending on the node’s density, in which one node is adaptively assigned as the Cluster Head (CH). A key novelty of the proposed EECIT lies in the routing stage where ranking sensor nodes is performed via a new metric named Fair Selection (FS) coefficient, which is designed by considering a combination of mean-deviation and the number of times that nodes involve in routing. As a consequence of the fair distribution of energy consumption across the network, the network’s lifetime increases.

With expected widespread implementation of 5G networks and 5G Internet of Things (IoT), indoor localization is expected to become of even further importance. Although Global Positioning System (GPS) ensures efficient outdoor localization, generally speaking, indoor localization systems fail to provide the same level of efficiency. In this regard, there has been recent widespread attention to Angle of Arrival (AoA) with the application on Switch Antenna Array (SAA), as an efficient indoor localization method due to its potential in determining location with low estimation error. The AoA, however, suffers from several issues including being sensitive to multipath effects, noise, fluctuations of received signal, and frequency/phase shifts. To tackle these issues, the paper proposes a set of signal processing and information fusion methods by integration of Nonlinear Least Square (NLS) curve fitting, Kalman Filter (KF), and Gaussian Filter (GF) to boost the accuracy rate of estimated angle. The proposed fusion framework is evaluated based on a real Bluetooth Low Energy (BLE) dataset and results illustrate significant potentials in terms of improving overall BLE-based achievable accuracy in angle detection.

Predicting death in lung cancer patients before initiating treatment is of paramount importance as this may guide decision-making towards more aggressive or combination of different types of treatment. In this work, we propose a Multi-scale Deep learning-based Radiomics model, referred to as "MDR-SURV" that exploits the information from positron emission tomography/computed tomography (PET/CT) images, combined with other clinical factors, to predict the overall survival (OS). Deep learning-based radiomics has the advantage of learning what features to extract, on its own. Furthermore, it does not require the exact segmentation of the tumor. The proposed MDR-SURV, which is a multi-scale framework, incorporates the tumor region and its surroundings, from different scales, and can extract both local and global tumor features. PET/CT images of 132 lung cancer patients who underwent stereotactic body radiotherapy (SBRT) were used to predict OS with the proposed model. Our results show that the MDR-SURV model outperforms its single-scale counterparts in predicting OS. Furthermore, the proposed MDR-SURV model achieves significantly high concordance index (C-index) of 73% in predicting the OS, which is noticeably higher than the results reported in existing literature.

With recent breakthroughs in signal processing, communication and networking systems, we are more and more surrounded by smart connected devices empowered by the Internet of Thing (IoT). Bluetooth Low Energy (BLE) is considered as the main-stream technology to perform identification and localization/tracking in IoT applications. Indoor localization applications within smart cities, typically, start by observing messages transmitted by BLE beacons and then utilization of Received Signal Strength Indicator (RSSI) to provide location estimates. RSSI signals are, however, prone to significant fluctuations. The main challenge is that multipath fading and drastic fluctuations in the indoor environment result in complex non-Gaussian RSSI measurements, necessitating the need to smooth RSSIs for development of BLE-based localization applications. In contrary to existing solutions, where RSSIs are assumed to have normal statistical properties, in this paper, a Gaussian Sum Filter (GSF) approach is designed to more realistically model the non-Gaussian nature of RSSIs. To maintain acceptable computational load, the number of components in the GSF is collapsed into a single Gaussian term with a novel Wasserstein Distance (WD)-Based Gaussian Mixture Reduction (GMR) algorithm. The simulation results based on real collected RSSI signals confirm the success of the proposed WD-based GSF framework compared to its conventional counterparts.

Angle of Arrival (AoA) approach with applications to Bluetooth Low Energy (BLE) has been recognized as an effective indoor localization method because of its ability for position determination with low estimation error. However, there are several issues including Carrier Frequency Offset (CFO), multipath effect, Inter-Symbol Interference (ISI), noise, and phase shifting faced by the AoA. To tackle these issues, we first highlight the wireless signal model in BLE standard and formulate the transmitted signal, wireless channel model, and the signal received by Linear Antenna Array (LAA). In addition, the paper introduces a novel fusion processing technique to eliminate the destructive impact of the wireless channel on the received signal, which leads to accurate angle detection following precise position estimation. The effectiveness of the proposed fusion processing method is evaluated through an experimental testbed in the presence of noise and Rayleigh fading channel. Based on the simulation results, the proposed processing approach illustrates significant improvements in the angle detection and path tracking in companion to its counterparts.

It is recognized that system trustworthiness is a hyperstructure embodied by the structural, behavioral, and system dimensions with a set of coherent attributes. We explore a theoretical framework of tripartite trustworthiness that can be applied to real-world autonomous systems. We present a formal study of the essences and mathematical models of system trustworthiness and their quantitative measurements in the contexts of autonomous and mission-critical intelligent systems where humans and machines interact in a hybrid environment.

This article proposes a Bayesian surprise minimization scheme to perform adaptive estimation for a family of linear Gaussian dynamic models. It is shown that the redefined Bayesian surprise in linear Gaussian dynamic systems is a function of the Kalman filter parameters and plays a key role in the state-estimation process. The proposed representation of the Kalman filter illustrates that the information from the Bayesian surprise and the innovation process contributes to the estimation of the state vector and its covariance matrix. This unique approach yields a new set of linear estimation algorithms, where filtering is purely performed with respect to the Bayesian surprise. Simulation results confirm that the information in Bayesian surprise can be sufficient to achieve optimal estimation. In addition, an alternative approach is proposed to test filter consistency based on Bayesian surprise.

The interconnectedness of all things is continuously expanding which has allowed every individual to increase their level of interaction with their surroundings. Internet of Things (IoT) devices are used in a plethora of context-aware application, such as proximity-based services (PBSs), and location-based services (LBSs). For these systems to perform, it is essential to have reliable hardware and predict a user's position in the area with high accuracy in order to differentiate between individuals in a small area. A variety of wireless solutions that utilize received signal strength indicators (RSSIs) have been proposed to provide PBS and LBS for indoor environments, though each solution presents its own drawbacks. In this article, Bluetooth low energy (BLE) beacons are examined in terms of their accuracy in proximity estimation. Specifically, a mobile application is developed along with three Bayesian filtering techniques to improve the BLE beacon proximity estimation accuracy. This includes a Kalman filter, a particle filter, and a nonparametric information (NI) filter. Since the RSSI is heavily influenced by the environment, experiments were conducted to examine the performance of beacons from three popular vendors in two different environments. The error is compared in terms of mean absolute error (MAE) and root mean squared error (RMSE). According to the experimental results, Bayesian filters can improve proximity estimation accuracy up to 30% in comparison with traditional filtering, when the beacon and the receiver are within 3 m.

In recent years, the Internet of Things (IoT) has grown to include the tracking of devices through the use of indoor positioning systems (IPSs) and location-based services (LBSs). When designing an IPS, a popular approach involves using wireless networks to calculate the approximate location of the target from devices with predetermined positions. In many smart building applications, LBS is necessary for efficient workspaces to be developed. In this article, we examine two memoryless positioning techniques,K-nearest neighbor (KNN) and Naive Bayes, and compare them with simple trilateration, in terms of accuracy, precision, and complexity. We present a comprehensive analysis between the techniques through the use of three popular IoT wireless technologies: 1) ZigBee; 2) Bluetooth low energy (BLE); and 3) WiFi (2.4-GHz band), along with three experimental scenarios to verify results across multiple environments. According to experimental results, KNN is the most accurate localization technique as well as the most precise. The received signal strength indicator data set of all the experiments is available online.

Convolutional neural networks (CNNs), which have been the state-of-the-art in many image-related applications, are prone to losing important spatial information between image instances. Capsule networks (CapsNets), on the other hand, are capable of leveraging such information through their routing by agreement process, making them powerful architectures for small datasets, such as medical imaging ones. Within the domain of medical imaging problems, brain tumor classification is of paramount importance, due to the deadly nature of this cancer and the consequences of the tumor misclassification. In our recent works, we showed potentials of developing CapsNet architecture for the task of brain tumor type classification. Similar to other deep learning models, however, CapsNets do not capture prediction uncertainty (coming from the uncertainty in the model weights, which is significantly important in keeping the human experts in the loop, by returning the uncertain samples. In this paper, we propose a Bayesian CapsNet framework, referred to as the BayesCap, that can provide not only the mean predictions, but also entropy as a measure of prediction uncertainty. Results show that filtering out the uncertain predictions can improve the accuracy, confirming that returning the uncertain predictions is an appropriate strategy for improving interpretability of the network.

In this paper, we propose a novel design of image deblurring in the form of one-shot convolution filtering that can directly convolve with naturally blurred images for restoration. The problem of optical blurring is a common disadvantage to many imaging applications that suffer from optical imperfections. Despite numerous deconvolution methods that blindly estimate blurring in either inclusive or exclusive forms, they are practically challenging due to high computational cost and low image reconstruction quality. Both conditions of high accuracy and high speed are prerequisites for high-throughput imaging platforms in digital archiving. In such platforms, deblurring is required after image acquisition before being stored, previewed, or processed for high-level interpretation. Therefore, on-the-fly correction of such images is important to avoid possible time delays, mitigate computational expenses, and increase image perception quality. We bridge this gap by synthesizing a deconvolution kernel as a linear combination of finite impulse response (FIR) even-derivative filters that can be directly convolved with blurry input images to boost the frequency fall-off of the point spread function (PSF) associated with the optical blur. We employ a Gaussian low-pass filter to decouple the image denoising problem for image edge deblurring. Furthermore, we propose a blind approach to estimate the PSF statistics for two Gaussian and Laplacian models that are common in many imaging pipelines. Thorough experiments are designed to test and validate the efficiency of the proposed method using 2054 naturally blurred images across six imaging applications and seven state-of-the-art deconvolution methods.

One of the challenges facing the adoption of digital pathology workflows for clinical use is the need for automated quality control. As the scanners sometimes determine focus inaccurately, the resultant image blur deteriorates the scanned slide to the point of being unusable. Also, the scanned slide images tend to be extremely large when scanned at greater or equal 20X image resolution. Hence, for digital pathology to be clinically useful, it is necessary to use computational tools to quickly and accurately quantify the image focus quality and determine whether an image needs to be re-scanned. We propose a no-reference focus quality assessment metric specifically for digital pathology images that operate by using a sum of even-derivative filter bases to synthesize a human visual system-like kernel, which is modeled as the inverse of the lens' point spread function. This kernel is then applied to a digital pathology image to modify high-frequency image information deteriorated by the scanner's optics and quantify the focus quality at the patch level. We show in several experiments that our method correlates better with ground-truth z -level data than other methods, which is more computationally efficient. We also extend our method to generate a local slide-level focus quality heatmap, which can be used for automated slide quality control, and demonstrate the utility of our method for clinical scan quality control by comparison with subjective slide quality scores.

In recent years, computer vision techniques have made large advances in image recognition and been applied to aid radiological diagnosis. Computational pathology aims to develop similar tools for aiding pathologists in diagnosing digitized histopathological slides, which would improve diagnostic accuracy and productivity amidst increasing workloads. However, there is a lack of publicly-available databases of (1) localized patch-level images annotated with (2) a large range of Histological Tissue Type (HTT). As a result, computational pathology research is constrained to diagnosing specific diseases or classifying tissues from specific organs, and cannot be readily generalized to handle unexpected diseases and organs. In this paper, we propose a new digital pathology database, the ``Atlas of Digital Pathology'' (or ADP), which comprises of 17,668 patch images extracted from 100 slides annotated with up to 57 hierarchical HTTs. Our data is generalized to different tissue types across different organs and aims to provide training data for supervised multi-label learning of patch-level HTT in a digitized whole slide image. We demonstrate the quality of our image labels through pathologist consultation and by training three state-of-the-art neural networks on tissue type classification. Quantitative results support the visually consistency of our data and we demonstrate a tissue type-based visual attention aid as a sample tool that could be developed from our database.

According to official statistics, cancer is considered as the second leading cause of human fatalities. Among different types of cancer, brain tumor is seen as one of the deadliest forms due to its aggressive nature, heterogeneous characteristics, and low relative survival rate. Determining the type of brain tumor has significant impact on the treatment choice and patient's survival. Human-centered diagnosis is typically error-prone and unreliable resulting in a recent surge of interest to automatize this process using convolutional neural networks (CNNs). CNNs, however, fail to fully utilize spatial relations, which is particularly harmful for tumor classification, as the relation between the tumor and its surrounding tissue is a critical indicator of the tumor's type. In our recent work, we have incorporated newly developed CapsNets to overcome this shortcoming. CapsNets are, however, highly sensitive to the miscellaneous image background. The paper addresses this gap. The main contribution is to equip CapsNet with access to the tumor surrounding tissues, without distracting it from the main target. A modified CapsNet architecture is, therefore, proposed for brain tumor classification, which takes the tumor coarse boundaries as extra inputs within its pipeline to increase the CapsNet's focus. The proposed approach noticeably outperforms its counterparts.

Recent advancements in signal processing and communication systems have resulted in evolution of an intriguing concept referred to as Internet of Things (IoT). By embracing the IoT evolution, there has been a surge of recent interest in localization/tracking within indoor environments based on Bluetooth Low Energy (BLE) technology. The basic motive behind BLE-enabled IoT applications is to provide advanced residential and enterprise solutions in an energy efficient and reliable fashion. Although recently different state estimation (SE) methodologies, ranging from Kalman filters, Particle filters, to multiple-modal solutions, have been utilized for BLE-based indoor localization, there is a need for ever more accurate and real-time algorithms. The main challenge here is that multipath fading and drastic fluctuations in the indoor environment result in complex non-linear, non-Gaussian estimation problems. The paper focuses on an alternative solution to the existing filtering techniques and introduces/discusses incorporation of the Belief Condensation Filter (BCF) for localization via BLE-enabled beacons. The BCF is a member of the universal approximation family of densities with performance bound achieving accuracy and efficiency in sequential SE and Bayesian tracking. It is a resilient filter in harsh environments where nonlinearities and non-Gaussian noise profiles persist, as seen in such applications as Indoor Localization.

In digital pathology, tissue slides are scanned into Whole Slide Images (WSI) and pathologists first screen for diagnostically-relevant Regions of Interest (ROIs) before reviewing them. Screening for ROIs is a tedious and time-consuming visual recognition task which can be exhausting. The cognitive workload could be reduced by developing a visual aid to narrow down the visual search area by highlighting (or segmenting) regions of diagnostic relevance, enabling pathologists to spend more time diagnosing relevant ROIs. In this paper, we propose HistoSegNet, a method for semantic segmentation of histological tissue type (HTT). Using the HTT-annotated Atlas of Digital Pathology (ADP) database, we train a Convolutional Neural Network on the patch annotations, infer Gradient-Weighted Class Activation Maps, average overlapping predictions, and post-process the segmentation with a fully-connected Conditional Random Field. Our method out-performs more complicated weakly-supervised semantic segmentation methods and can generalize to other datasets without retraining.

Brain tumor, which is one of the deadliest cancers, can have several types, based on different characteristics of the tumor. Determining the exact category of this cancer is of significant importance, because it directly affects the treatment options and patient's survival. Although experts' judgment remains the gold standard for brain tumor classification, human-centered decision is time-consuming and error prone. Radiomics, referring to the extraction of features from medical images with the ultimate goal of cancer prediction/classification, can automatize the tumor classification task, having the promise of providing more accurate and time-effective decision. Hand-crafted Radiomics, which is the most common Radiomics analysis, needs a prior knowledge on the types of features to extract, which is not always available, leading to an increased surge of interest toward deep learning-based Radiomics. Although deep learning-based Radiomics does not need a prior knowledge, its practical application is limited by the low explainablity of the deep learning networks. In this work, the interpretability of Capsule networks, which have shown promising results in brain tumor classification, is explored. Outcomes show that the Radiomics features extracted by Capsule networks can not only distinguish between the tumor types, but also show considerable correlation with hand-crafted features, which are more acceptable and reliable from a physician's point of view.

With the rapid emergence of Internet of Things (IoT), we are more and more surrounded by smart connected devices with integrated sensing, processing, and communication capabilities. Bluetooth Low Energy (BLE), referred to as Bluetooth Smart, is considered as the main-stream technology to perform identification and localization/tracking in IoT applications. While single-model BLE-based tracking has been investigated from different aspects, application of multi-model (hybrid) solutions are still in their infancy. In this regard, the letter proposes a novel BLE-based tracking framework, referred to as the STUPEFY, which incorporates set-valued information within box particle filtering context. More specifically, the proposed multiple-model STUPEFY framework consists of three integrated modules, i.e., an intriguing Smoothing Module based on Kalman filtering to reduce the Received Signal Strength Indicator (RSSI) fluctuations and facilitate comparison of Gaussian models of the RSSI values in distribution with the learned ones; A learning-based model (Coordination Module) utilized in an intuitive fashion to provide/construct a coarse estimate of the target's location together with the smallest axes-aligned box containing the ellipsoid associated with each zone's learned RSSI distribution, and; A novel set-valued box particle filtering (SBPF) approach (Micro-Localization Module). The proposed STUPEFY framework is evaluated based on real BLE datasets and results illustrate significant potentials in terms of improving overall BLE-based achievable tracking accuracy.

This paper explores the influence of high cognitive load on vehicle driver's electroencephalography (EEG) signals collected from two channels (Fp1, Fp2) using a wireless consumer-grade system. Although EEG has been used in driving-related research to assess cognitive load, only a few studies focused on high load, and they used research-grade systems. Recent advancements allow for less intrusive and more affordable systems. As an exploration, we tested the feasibility of one such system to differentiate among three levels of cognitive taskload in a simulator study. Thirty-seven participants completed a baseline drive with no secondary task and two drives with a modified version of the n-back task (1-back and 2-back). The modification removed the verbal response required during task presentation to prevent EEG-signal degradation, with the 2-back task expected to impose higher load than that by the 1-back task. Another objective of this study is to validate that this modified task increased the cognitive load in the expected manner. The modified task led to significant trends from baseline to 1-back, and from 1-back to 2-back in participants' heart rate, galvanic skin response, respiration, horizontal gaze position variability, and pupil diameter, all in line with the previous driving-related studies on cognitive load. Furthermore, the EEG system was observed to be sensitive to the modified task, with the power of alpha band decreasing significantly with increasing n-back levels (baseline versus 1-back: 0.092 Bels on Fp1, 0.179 on Fp2; 1-back versus 2-back: 0.209 on Fp1, 0.147 on Fp2). Thus, a consumer-grade EEG system has the potential to capture high levels of cognitive load experienced by drivers.

Touchless palmprint recognition systems enable high-accuracy recognition of individuals through less-constrained and highly usable procedures that do not require the contact of the palm with a surface. To perform this recognition, methods based on local texture descriptors and convolutional neural networks (CNNs) are currently used to extract highly discriminative features while compensating for variations in scale, rotation, and illumination in biometric samples. In particular, the main advantage of CNN-based methods is their ability to adapt to biometric samples captured with heterogeneous devices. However, the current methods rely on either supervised training algorithms, which require class labels (e.g., the identities of the individuals) during the training phase, or filters pretrained on general-purpose databases, which may not be specifically suitable for palmprint data. To achieve a high-recognition accuracy with touchless palmprint samples captured using different devices while neither requiring class labels for training nor using pretrained filters, we introduce PalmNet, which is a novel CNN that uses a newly developed method to tune palmprint-specific filters through an unsupervised procedure based on Gabor responses and principal component analysis (PCA), not requiring class labels during training. PalmNet is a new method of applying Gabor filters in a CNN and is designed to extract highly discriminative palmprint-specific descriptors and to adapt to heterogeneous databases. We validated the innovative PalmNet on several palmprint databases captured using different touchless acquisition procedures and heterogeneous devices, and in all cases, a recognition accuracy greater than that of the current methods in this paper was obtained.

In this paper, we propose a novel design of Human Visual System (HVS) response in a convolutional filter form to decompose meaningful features that are closely tied with image sharpness level. No-reference (NR) Image sharpness assessment (ISA) techniques have emerged as the standard of image quality assessment in diverse imaging applications. Despite their high correlation with subjective scoring, they are challenging for practical considerations due to high computational cost and lack of scalability across different image blurs. We bridge this gap by synthesizing the HVS response as a linear combination of Finite Impulse Response (FIR) derivative filters to boost the falloff of high band frequency magnitudes in natural imaging paradigm. The numerical implementation of the HVS filter is carried out with MaxPol filter library that can be arbitrarily set for any differential orders and cutoff frequencies to balance out the estimation of informative features and noise sensitivities. Utilized by HVS filter, we then design an innovative NR-ISA metric called "HVS-MaxPol" that (a) requires minimal computational cost, (b) produce high correlation accuracy with image sharpness level, and (c) scales to assess synthetic and natural image blur. Specifically, the synthetic blur images are constructed by blurring the raw images using Gaussian filter, while natural blur is observed from real-life application such as motion, out-of-focus, luminance contrast, etc. Furthermore, we create a natural benchmark database in digital pathology for validation of image focus quality in whole slide imaging systems called "FocusPath" consisting of 864 blurred images. Thorough experiments are designed to test and validate the efficiency of HVS-MaxPol across different blur databases and state-of-the-art NR-ISA metrics. The experiment result indicates that our metric has the best overall performance with respect to speed, accuracy and scalability.

This paper introduces a computer-aided kidney shape detection method suitable for volumetric (3D) ultrasound images. Using shape and texture priors, the proposed method automates the process of kidney detection, which is a problem of great importance in computer-assisted trauma diagnosis. This paper introduces a new complex-valued implicit shape model, which represents the multiregional structure of the kidney shape. A spatially aligned neural network classifiers with complex-valued output is designed to classify voxels into background and multiregional structure of the kidney shape. The complex values of the shape model and classification outputs are selected and incorporated in a new similarity metric, such as the shape-to-volume registration process only fits the shape model on the actual kidney shape in input ultrasound volumes. The algorithm's accuracy and sensitivity are evaluated using both simulated and actual 3-D ultrasound images, and it is compared against the performance of the state of the art. The results support the claims about accuracy and robustness of the proposed kidney detection method, and statistical analysis validates its superiority over the state of the art.

The paper is motivated by recent advancements and developments in large, distributed, autonomous, and self-aware systems such as autonomous vehicles and vehicle-to-everything (V2X) technologies, where bandwidth, security, privacy, and/or power considerations limit the number of information transfers between neighbouring agents. In this regard, we propose an event-triggered distributed state estimation via diffusion strategies (ET/DPF), which is a systematic and intuitively pleasing distributed state estimation algorithm that jointly incorporates point and set-valued measurements within the particle filtering framework. In the absence of a measurement form a neighbouring node (i.e., having a set-valued measurement), each local agent/node evaluates the probability that the unknown measurement belongs to the event-triggering set based on its particles which is then used to update the corresponding particle weights. In our Monte Carlo simulations, the proposed ET/DPF outperforms its counterparts in environments with limited bandwidth or/and intermittent connectivity.

Brain tumor is considered as one of the deadliest and most common form of cancer both in children and in adults. Consequently, determining the correct type of brain tumor in early stages is of significant importance to devise a precise treatment plan and predict patient's response to the adopted treatment. In this regard, there has been a recent surge of interest in designing Convolutional Neural Networks (CNNs) for the problem of brain tumor type classification. However, CNNs typically require large amount of training data and can not properly handle input transformations. Capsule networks (referred to as CapsNets) are brand new machine learning architectures proposed very recently to overcome these shortcomings of CNNs, and posed to revolutionize deep learning solutions. Of particular interest to this work is that Capsule networks are robust to rotation and affine transformation, and require far less training data, which is the case for processing medical image datasets including brain Magnetic Resonance Imaging (MRI) images. In this paper, we focus to achieve the following four objectives: (i) Adopt and incorporate CapsNets for the problem of brain tumor classification to design an improved architecture which maximizes the accuracy of the classification problem at hand; (ii) Investigate the over-fitting problem of CapsNets based on a real set of MRI images; (iii) Explore whether or not CapsNets are capable of providing better fit for the whole brain images or just the segmented tumor, and; (iv) Develop a visualization paradigm for the output of the CapsNet to better explain the learned features. Our results show that the proposed approach can successfully overcome CNNs for the brain tumor classification problem.

The paper is motivated by the fact that brain cancer is one of the deadliest cancers and its detection in early stages is of paramount importance. In this regard, tumor 3D shape reconstruction from magnetic resonance (MR) or computed tomography (CT) scans provides critical information, which can not be interpreted from 2D images. However, CT and MR images have low resolution in z direction compared to their resolution in x and y directions, therefore, 3D reconstructed shapes are of low quality. In this paper, we propose to use convolutional auto-encoders (CAEs) to address this drawback, and develop a convolutional autoencoder-based inter-slice interpolation (CARISI) framework. Although deep nets have been used very recently for brain tumor segmentation, to the best of our knowledge, this is the first attempt to use CAEs for 3D reconstruction of brain tumor. The proposed CARISI framework consists of several encoding and decoding components, which can handle rapid changes in tumor shape without the need for supervision of an expert. Our experiments based on a real data-set consisting of 3064 segmented brain tumor images indicate that the proposed CARISI framework outperforms its counterpart and has the potential to significantly improve the overall quality of the reconstructed shapes.

We presents a no-reference (NR) image sharpness metric based on a visual sensitivity model. We propose that MaxPol convolution kernels are close approximation to this model and capable of extracting meaningful features for image sharpness assessment. Equipped by these kernels, we develop an efficient pipeline to evaluate the out-of-focus level of input images by decomposing the first and third order image differentials. The associated kernels are regulated in higher cutoff frequencies to balance out the information loss and noise sensitivity. We use high order central moments to exploit sharpness scores in wide range of frequency information. The experimental results outperform the state-of-the-art methods in accuracy and speed.

This letter proposes a novel ternary-event-based particle filtering (TEB-PF) framework by introducing the ternary-event-triggering mechanism coupled with a non-Gaussian fusion strategy that jointly incorporates point-valued, quantized, and set-valued measurements. In contrast to the existing binary-event-triggering solutions, the TEB-PF is a distributed state estimation architecture where the remote sensor communicates its measurements to the estimator, residing at the fusion centre, in a ternary-event-based fashion, i.e., holds on to its observation during idle epochs, transfers quantized ones during the transitional epochs, and only communicates raw observations during event epochs. Due to joint utilization of quantized and set-valued measurements in addition to the point-valued ones, the proposed TEB-PF simultaneously reduces the communication overhead, in comparison to its binary triggering counterparts, while also improving the estimation accuracy specially in low communication rates.

Gaussian mixtures (GMs) are widely used in signal processing applications to capture the multimodal behavior of dynamic systems. Due to an exponential increase in the number of GM components in such applications, Gaussian mixture reduction (GMR) approaches are deemed necessary. Traditionally, the Kullback-Leibler divergence (KLD) is used for GMR along with a moment-matching merging approach to minimize the information loss. However, in certain applications such as image retrieval, preserving the geometric shape of the GM is more appealing. For such applications, this work prescribes the use of the Wasserstein distance (WD), which quantifies the minimum cost of converting one density into another and, therefore, is mostly concerned with the shape difference between the densities. To minimize the change in the shape of the GM, first, similar GM components are identified utilizing the WD. Next, these components are merged by proposing a novel WD-based averaging method. The simulation results confirm the success of the proposed WD-based GMR techniques in providing a better approximation of the original GM in the WD sense as compared to KLD-based methods.

In a smart home, the home status, as well as the human activities, can be observed through a number of sensors. A wireless network can transfer the data to an information system and the commands from the information system to the sensors and actuators. In small areas such as smart homes, four types of noise may form in communication system. In this work, we explore how the noise can be resolved by integrating the Wireless Sensor Network (WSN) management with the smart home information system. The result is to smoothen the wireless communication. Additionally, the sensors and actuators are applied efficiently, since the information system places them automatically in the right place at right time. The proposed Opportunistic Mesh (OPM) wireless method avoids signal interference/ collision in small rooms, in order to minimize interference with home regular wireless services, such as WiFi. A model to vector the mobile agent in the smart home is proposed and a mobile agent is used to automatically approach the target positions.

A systematic and comprehensive framework for finite impulse response (FIR) lowpass/fullband derivative kernels is introduced in this paper. Closed form solutions of a number of derivative filters are obtained using the maximally flat technique to regulate the Fourier response of undetermined coefficients. The framework includes arbitrary parameter control methods that afford solutions for numerous differential orders, variable polynomial accuracy, centralized/staggered schemes, and arbitrary side-shift nodes for boundary formulation. Using the proposed framework, four different derivative matrix operators are introduced and their numerical stability is analyzed by studying their eigenvalues distribution in the complex plane. Their utility is studied by considering two important image processing problems, namely gradient surface reconstruction and image stitching. Experimentation indicates that the new derivative matrices not only outperform commonly used methods but provide useful insights to the numerical issues in these two applications.

It is well known that the performance of the Kalman filter deteriorates when the system noise statistics are not available a priori. In particular, the adjustment of measurement noise covariance is deemed paramount as it directly affects the estimation accuracy and plays the key role in applications such as sensor selection and sensor fusion. This letter proposes a novel adaptive scheme by approximating the measurement noise covariance distribution through finite samples, assuming the noise to be white with a normal distribution. Exploiting these samples in approximation of the system state a posteriori leads to a Gaussian mixture model (GMM), the components of which are acquired by Kalman filtering. The resultant GMM is then reduced to the closest normal distribution and also used to estimate the measurement noise covariance. Compared to previous adaptive techniques, the proposed method adapts faster to the unknown parameters and thus provides a higher performance in terms of estimation accuracy, which is confirmed by the simulation results.

A generalized framework for numerical differentiation (ND) is proposed for constructing a finite impulse response (FIR) filter in closed form. The framework regulates the frequency response of ND filters for arbitrary derivative-order and cutoff frequency selected parameters relying on interpolating power polynomials and maximally flat design techniques. Compared with the state-of-the-art solutions, such as Gaussian kernels, the proposed ND filter is sharply localized in the Fourier domain with ripple-free artifacts. Here, we construct 2D MaxFlat kernels for image directional differentiation to calculate image differentials for arbitrary derivative order, cutoff level and steering angle. The resulted kernel library renders a new solution capable of delivering discrete approximation of gradients, Hessian, and higher-order tensors in numerous applications. We tested the utility of this library on three different imaging applications with main focus on the unsharp masking. The reported results highlight the high efficiency of the 2D MaxFlat kernel and its versatility with respect to robustness and parameter control accuracy.

In digital pathology, to address color variation and histological component colocalization in pathology images, stain decomposition is usually performed preceding spectral normalization and tissue component segmentation. This paper examines the problem of stain decomposition, which is a naturally nonnegative matrix factorization (NMF) problem in algebra, and introduces a systematical and analytical solution consisting of a circular color analysis module and an NMF-based computation module. Unlike the paradigm of existing stain decomposition algorithms where stain proportions are computed from estimated stain spectra using a matrix inverse operation directly, the introduced solution estimates stain spectra and stain depths via probabilistic reasoning individually. Since the proposed method pays extra attentions to achromatic pixels in color analysis and stain co-occurrence in pixel clustering, it achieves consistent and reliable stain decomposition with minimum decomposition residue. Particularly, aware of the periodic and angular nature of hue, we propose the use of a circular von Mises mixture model to analyze the hue distribution, and provide a complete color-based pixel soft-clustering solution to address color mixing introduced by stain overlap. This innovation combined with saturation-weighted computation makes our study effective for weak stains and broad-spectrum stains. Extensive experimentation on multiple public pathology datasets suggests that our approach outperforms state-of-the-art blind stain separation methods in terms of decomposition effectiveness.

Automated segmentation of kidneys in three-dimensional (3-D) abdominal ultrasound volumes is a task of paramount importance in automated diagnosis of abdominal trauma. However, ultrasound speckle noise, low-contrast boundaries, partial kidney occlusion, and probe misalignment restrict the utility of the solution, especially when it is used in emergency rooms and Focused Assessment with Sonography Trauma applications. This paper introduces a systematic and cost-effective method capable of detecting and segmenting the kidney's shape in acquired 3-D ultrasound volumes, using off-line training datasets. This paper offers a new shape model representation, called the complex-valued implicit shape model, to generate a 3-D kidney shape model by combining prior knowledge of training shapes and anatomical knowledge. We apply shape-to-volume registration, based on a new similarity metric, to detect the kidney shape by fitting the 3-D shape model on 3-D ultrasound volumes. Upon kidney detection, the fitted shape model is used to initialize and evolve a new level-set function, called complex-valued rational level-set with shape prior, to segment the kidney's shape. Experimentation using both simulated and actual ultrasound volumes indicate that the proposed solution provides a better performance over the state-of-the-art volumetric ultrasound segmentation methods.

Texture is considered one of most significant information sources in histo-pathology image analysis. To take advantage of information on color texture in digital histo-pathology, this work analyzes inherent characteristics of the hue component in the cylindrical color space, and introduces an effective color texture descriptor based on the LBP paradigm. Unlike existing LBP variants designed for linear data, the proposed descriptor, namely Hue-LBP, addresses the angular and periodic nature of hue and shows that color variation in the hue channel can be quantified by an angular variable in the range of [0,180]. By introducing the concept of color similarity as a metric to measure color variation, we obtain a histogram to describe local color texture patterns. Experimentation on histo-pathology image classification suggests that the proposed Hue-LBP is discriminative as it is capable of describing texture information conveyed by the hue components.

This paper introduces a segmentation approach, where a discriminative dictionary with objects' shape information is learned, followed by a sparse representation based segmentation process. In contrast with state-of-the-art sparse representation classification methods using discriminative dictionary learning, the proposed method learns a discriminative dictionary containing both intensity and shape information of object classes, in which shape information is collected and represented in the form of binarized masks. Object segmentation is achieved through an iterative process, including sparse representation, shape estimation, and shape refinement. The introduced method is evaluated and compared to state-of-the-art sparse representation based segmentation methods, and demonstrated better segmentation performance.

Affective states of a user provide important information for many applications such as, personalized information (e.g., multimedia content) retrieval/delivery or intelligent human-computer interface design. In recently years, physiological signals, Electroencephalogram (EEG) in particular, have been shown to be very effective in estimating a user's affective states during social interaction or under video or audio stimuli. However, due to the large number of parameters associated with the neural expression of emotion, there is still a lot of unknowns on the specific spatial and spectral correlation of the EEG signal and the affective states expression. To investigate on such correlation, two types of semi-supervised deep learning approaches, stacked denoising autoencoder (SDAE) and deep belief networks (DBN), were applied as application specific feature extractors for the affective states classification problem using EEG signals. To evaluate the efficacy of the proposed semi-supervised approaches, a subject-specific affective states classification experiment were carried out on the DEAP database to classify 2-dimensional affect states. The DBN based model achieved averaged F1 scores of 86.67%, 86.60% and 86.69% for arousal, valence and liking states classification respectively, which has significantly improved the state-of-art classification performance. By examining the weight vectors at each layer, we were also able to gain insights on the spatial or spectral locations of the most discriminating features. Another main advantage of applying the semi-supervised learning methods is that only a small fraction of labeled data, e.g., 1/6 of the training samples, were used in this study.

In this paper, a new computer-assisted ultrasound probe placement system is introduced to guide paramedics and first responders to conduct abdominal ultrasound imaging for diagnosing trauma patients in emergency situations where specialists are not present. Recently, telesonography has been employed to supervise paramedics by remote experts to perform ultrasound scan for triaging, although its utility is limited by unavailability of fast internet connectivity in remote regions. In the proposed solution of this paper, a paramedic is first instructed to place the ultrasound probe on an initial placement for imaging an organ of interest. Then, a three-dimensional (3-D) ultrasound image is acquired and processed to determine the organ's shape misalignment with respect to a reference alignment. Afterward, the organ's shape misalignment is used to estimate the probe misalignment, and then, a probe placement command is generated to guide the paramedic. This process iterates until a correct organ's view of interest is obtained. As the advantage of the proposed solution over the existing technique, the proposed solution does not require a fast Internet connectivity and a dedicated remote specialist to conduct ultrasound imaging by a paramedic. The utility of the proposed solution is evaluated for a case study on the right upper quadrant (RUQ) view, which has a paramount importance in triaging trauma patients. Accuracy and robustness of the proposed solution is verified using actual and simulated 3-D ultrasound images of the RUQ view.

Analysis of the statistical properties of natural images has played a vital role in the design of no-reference (NR) image quality assessment (IQA) techniques. In this paper, we propose parametric models describing the general characteristics of chromatic data in natural images. They provide informative cues for quantifying visual discomfort caused by the presence of chromatic image distortions. The established models capture the correlation of chromatic data between spatially adjacent pixels by means of color invariance descriptors. The use of color invariance descriptors is inspired by their relevance to visual perception, since they provide less sensitive descriptions of image scenes against viewing geometry and illumination variations than luminances. In order to approximate the visual quality perception of chromatic distortions, we devise four parametric models derived from invariance descriptors representing independent aspects of color perception: 1) hue; 2) saturation; 3) opponent angle; and 4) spherical angle. The practical utility of the proposed models is examined by deploying them in our new general-purpose NR IQA metric. The metric initially estimates the parameters of the proposed chromatic models from an input image to constitute a collection of quality-aware features (QAF). Thereafter, a machine learning technique is applied to predict visual quality given a set of extracted QAFs. Experimentation performed on large-scale image databases demonstrates that the proposed metric correlates well with the provided subjective ratings of image quality over commonly encountered achromatic and chromatic distortions, indicating that it can be deployed on a wide variety of color image processing problems as a generalized IQA solution.

Motivated by application of complex-valued signal processing techniques in statistical pattern recognition, classification, and Gaussian mixture (GM) modeling, this paper derives analytical expressions for computing the Bhattacharyya coefficient/distance (BC/BD) between two improper complex-valued Gaussian distributions. The BC/BD is one of the most widely used statistical measures for evaluating class separability in classification problems, feature extraction in pattern recognition, and for GM reduction (GMR) purposes. The BC provides an upper bound on the Bayes error, which is commonly known as the best criterion to evaluate feature sets. Although the computation of the BC/BD between real-valued signals is a well-known result, it has not yet been extended to the case of improper complex-valued Gaussian densities. This paper addresses this gap. We analyze the role of the pseudocovariance matrix, which characterizes the noncircularity of the signal, and show that it carries critical second-order statistical information for computing the BC/BD. We derive upper and lower bounds on the BD in terms of the eigenvalues of the covariance and pseudocovariance matrices of the underlying densities. The theoretical bounds are then used to introduce the concept of β -dominance in the context of statistical distance measures. The BC is pseudometric, since it fails to satisfy the triangle inequality. Using the Matusita distance (a full-metric variant of the BC), we propose an intuitively pleasing indirect distance measure for comparing two general GMs. Finally, we investigate the application of the proposed BC/BD measures for GMR purposes and develop two BC-based GMR algorithms.

In this paper, we present a novel approach to achieve blind stain decomposition in histo-pathology images. The method is based on stain color estimation, followed by stain absorbing vector generation and matrix computation. Unlike conventional approaches adopting linear processing algorithms to analyze chromatic information in the cylindrical-coordinate color spaces, which may be inappropriate for circular data such as hue, we propose the use of circular thresholding on saturation-weighted hue histogram to compute candidates for stain representative colors. Experimental results suggest that our stain decomposition method is capable to address spectral variation in stains effectively. We compare the proposed method to state-of-the-art blind stain separation algorithms for nuclei segmentation on breast histo-pathology images, and demonstrate that the segmentation scheme adopting our method in its pre-processing step achieves the best segmentation results.

This paper introduces a spectrum sensing scenario based on kernel theory which compares favorably against the conventional Energy Detector (ED) in a cognitive radio system. The so-called Kerenlized Energy Detector (KED) can provide superior accuracy in the case of non-Gaussian noise. The incorporation of the nonlinear kernel function in the KED test statistics allows for the development of a nonlinear algorithm capable of considering both higher order and Fractional Lower Order Moments (FLOMs) in the sensing task. Simulation results show that the proposed semi-blind kernelized spectrum sensing algorithm is much robust against impulsive noises and displays a considerably better detection performance than the conventional ED in practical impulsive man-made noises which are generally modeled as the Laplacian and the α-stable distributions. Moreover, for the Gaussian signal and noise model, the performance of the KED scheme is almost identical to that of the conventional ED.

Sparse representation-based classification (SRC) has been recently attracted a great interest among the signal processing society. SRC applies a discriminative representation using training samples to separate signals into their classes. In existing SRC methods, the dictionary size, which highly affects the performance, is manually set. Moreover, they are linear classifiers, and thus, they are not suitable for classifying nonlinear problems. In this paper, we propose a new classification method by cascading a dictionary learning and the neural network to take the advantages of both methods. We use dictionary learning with efficient number of elements (DLENE) to extract discriminative features. We also use the proposed binomial classifier to detect kidneys in 3D ultrasound images. A set of Caltech-101 images are used to compare the proposed method with the state-of-the-art. The proposed kidney detection is evaluated by a set of ultrasound volumes. The results confirm the superiority of our proposed method.

The conventional numerical approaches in wide range of restoration problems encode either first or second order total variation (TV) using low accuracy FIR filters, i.e. [−1, 1] and [1, −2, 1]. This leads to inappropriate feature estimation of high-frequency components in an underlying signal for reconstruction. We introduce high-order-of-accuracy (HOA) numerical differentiation to encode such features in combined first and second order TV regularization. In particular we design appropriate gradient and hessian operators embedded with HOA filters to incorporate in combined regularizers. We seek the solution to the combined approach using the alternating direction methods of multipliers minimization algorithm to reconstruct three dimensional sparse tensors. Particular application of this combined regularizer is studied over compressed video sensing problem. Numerical experiments state significantly better recoveries over consecutive frames from their compressed measurements.

In this paper, we present a novel approach to achieve diagnostic color estimation for histological objects in pathology images. The method is based on a von Mises mixture model for hue histogram, followed by implicit pixel clustering via maximum likelihood estimation and representative color computation. Unlike conventional approaches adopting linear processing algorithms to analyze hue histogram which is characterized by a nature of periodicity, we build a circular cluster model composed of multiple von Mises distributions to address the directional nature of hue. Experimental results on synthetic circular data suggest that the proposed circular model outperforms both classical linear thresholding methods and the state-of-art circular thresholding approach in terms of cluster parameter estimation. The color estimation experiment on publicly-accessible cytopathology images demonstrates that our method is capable to accurately estimate object's diagnostic color, which can be used for subsequent image analysis.

The letter considers a multi-sensor state estimation problem configured in a decentralized architecture where local complex statistics are communicated to the central processing unit for fusion instead of the raw observations. Naive adaptation of the augmented complex statistics to develop a decentralized state estimation algorithm results in increased local computations, and introduces extensive communication overhead, making it practically unattractive. The letter proposes a structure-induced complex Kalman filter framework with reduced communication overhead. In order to further reduce the local computations, the letter proposes a non-circularity criterion which allows each node to examine the non-circularity of its local observations. A local sensor node disregards its extra second-order statistical information when the non-circularity coefficient is small. In cases where the local observations are highly non-circular, an intuitively pleasing circularization approach is proposed to avoid computation and communication of the pseudo-covariance matrices. Simulation results indicate that the proposed structured-induced complex Kalman filter (SCKF) provides significant performance improvements over its traditional counterparts.

Motivated by application of Gaussian sum filters (GSF) and multiple model adaptive estimation (MMAE) approaches in scenarios where assumption of proper (circular) Gaussian signals is not valid, the letter proposes a novel complex-valued Gaussian sum filter (C/GSF) for non-linear filtering of non-Gaussian/non-circular measurement noise. Although the literature on recursive state estimation using GSF is rich, its complex-valued counterpart which incorporates the full second-order statistics of the system and can cope with non-Gaussian/non-circular measurements, has not yet been investigated in the literature. The paper addresses this gap. The C/GSF is a computationally attractive adaptive filter where the number of non-circular Gaussian components is controlled utilizing a modified Bayesian learning technique which is used to collapse the resulting non-Gaussian sum mixture into an equivalent complex-valued Gaussian term. Simulation results indicate that the C/GSF provides significant performance improvement over its traditional counterparts.

This paper presents a novel computational model for quantifying the perceptual quality of color images consistently with subjective evaluations. The proposed full-reference color metric, namely, a directional statistics-based color similarity index, is designed to consistently perform well over commonly encountered chromatic and achromatic distortions. In order to accurately predict the visual quality of color images, we make use of local color descriptors extracted from three perceptual color channels: 1) hue; 2) chroma; and 3) lightness. In particular, directional statistical tools are employed to properly process hue data by considering their periodicities. Moreover, two weighting mechanisms are exploited to accurately combine locally measured comparison scores into a final score. Extensive experimentation performed on large-scale databases indicates that the proposed metric is effective across a wide range of chromatic and achromatic distortions, making it better suited for the evaluation and optimization of color image processing algorithms.

This paper introduces a novel metric for image difference prediction, capable of handling color data. The proposed metric, namely, color difference index based on circular hue, is a full-reference based scheme, which independently processes achromatic and chromatic differences of two input color images. Within the framework, chromatic information is analyzed using two perceptual attributes, hue and chroma information, simulating human visual system mechanism. Unlike conventional approaches where the periodic nature of hue is disregarded, we propose to estimate hue difference by adopting theory of circular statistics. Performance of the proposed solution is validated using benchmark image quality assessment databases. Experimental results indicate the effectiveness of the proposed metric against a wide range of distortions, especially on chromatic distortions, making it better suited for color gamut mapping applications.

In classification of matrix-variate data, two-directional linear discriminant analysis (2DLDA) methods extract discriminant features while preserving and utilizing the matrix structure. These methods provide computational efficiency and improved performance in small sample size problems. Existing 2DLDA solutions produce a feature matrix which is commonly vectorized for processing by conventional vector-based classifiers. However, the vectorization step requires a one-dimensional ranking of features according to their discriminance power. We first demonstrate that independent column-wise and row-wise ranking provided by 2DLDA is not sufficient for uniquely sorting the resulting features, and does not guarantee the selection of the most discriminant features. Then, we theoretically derive the desired global ranking score based on Fisher's criterion. The current results focus on non-iterative solutions, but future extensions to iterative 2DLDA variants are possible. Face recognition experiments using images from the PIE data set are used to demonstrate the theoretically proved improvements over the existing solutions.

In this paper, we propose a new color difference assessment scheme, with emphasis on hue attribute. The proposed measure is a full-reference difference predictor, which extracts hue histogram descriptors from a pair of color images to be compared and quantifies their difference. In order to achieve good correlation with subjective judgement, we exploit a preprocessor approximating the contrast sensitivity function of visual system, followed by a histogram distance function for circular data which corresponds to the way humans perceive difference. Performance of the proposed scheme is validated on benchmark datasets exhibiting perceptual degradation caused by gamut mapping. Experimental results indicate the effectiveness of the proposed metric against chromatic distortions, making it useful as an optimization module for color gamut mapping applications.

The authors propose Markovian‐based spectrum sensing policies in a cognitive radio system that leverages past sensing outcomes of several cooperating secondary users (SUs) to decide which channel (of primary users – PUs) should be sensed by each SU at a given time. These policies are based on a new finite‐state channel model that captures the fading condition as well as the occupancy state for each primary channel. The multiuser extension of this model is useful when multiple spatially distributed SUs share their sensing outcomes. The proposed schemes allow the asynchronous sensing outcomes obtained by the SUs over different slots to be fused together and converted into a posteriori probabilities for the current states of the primary channels. As the detection threshold in a spectrum detector balances the trade‐off between the false‐alarm and miss probabilities for detecting primary signals in a single primary channel, a design parameter is introduced to allow the system designer to devise policies with different levels of aggressiveness. The authors evaluate the optimality and complexity of the proposed sensing policies and show that our schemes significantly increase secondary use of the spectrum and/or reduce interference with PUs compared to a random selection policy or a cooperative sensing policy based on a two‐state channel model.

Automatic facial expression recognition (FER) is becoming increasingly important in the area of affective computing systems because of its various emerging applications such as human-machine interface and human emotion analysis. Recently, sparse representation based FER has become popular and has shown an impressive performance. However, sparse representation could often produce less meaningful sparse solution for FER due to intra-class variation such as variation in identity or illumination. This paper proposes a new sparse representation based FER method, aiming to reduce the intra-class variation while emphasizing the facial expression in a query face image. To that end, we present a new method for generating an intra-class variation image of each expression by using training expression images. The appearance of each intra-class variation image could be close to the appearance of the query face image in identity and illumination. Therefore, the differences between the query face image and its intra-class variation images are used as the expression features for sparse representation. Experimental results show that the proposed FER method has high discriminating capability in terms of improving FER performance. Further, the intra-class variation images of non-neutral expressions are complementary with that of neutral expression, for improving FER performance.

Numerous total variation (TV) regularizers, engaged in image restoration problem, encode the gradients by means of simple [-1, 1] finite-impulse-response (FIR) filter. Despite its low computational processing, this filter severely distorts signal's high-frequency components pertinent to edge/ discontinuous information and cause several deficiency issues known as texture and geometric loss. This paper addresses this problem by proposing an alternative model to the TV regularization problem via high-order accuracy differential FIR filters to preserve rapid transitions in signal recovery. A numerical encoding scheme is designed to extend the TV model into multidimensional representation (tensorial decomposition). We adopt this design to regulate the spatial and temporal redundancy in compressed video sensing problem to jointly recover frames from undersampled measurements. We then seek the solution via alternating direction methods of multipliers and find a unique solution to quadratic minimization step with capability of handling different boundary conditions. The resulting algorithm uses much lower sampling rate and highly outperforms alternative state-of-the-art methods. This is evaluated both in terms of restoration accuracy and visual quality of the recovered frames.

Recent achievements in wireless technologies have opened up enormous opportunities for the implementation of ubiquitous health care systems in providing rich contextual information and warning mechanisms against abnormal conditions. This helps with the automatic and remote monitoring/tracking of patients in hospitals and facilitates and with the supervision of fragile, elderly people in their own domestic environment through automatic systems to handle the remote drug delivery. This paper presents a new modeling and analysis framework for the multipatient positioning in a wireless body area network (WBAN) which exploits the spatial sparsity of patients and a sparse fast Fourier transform (FFT)-based feature extraction mechanism for monitoring of patients and for reporting the movement tracking to a central database server containing patient vital information. The main goal of this paper is to achieve a high degree of accuracy and resolution in the patient localization with less computational complexity in the implementation using the compressive sensing theory. We represent the patients' positions as a sparse vector obtained by the discrete segmentation of the patient movement space in a circular grid. To estimate this vector, a compressive-sampling-based two-level FFT (CS-2FFT) feature vector is synthesized for each received signal from the biosensors embedded on the patient's body at each grid point. This feature extraction process benefits in the combination of both short-time and long-time properties of the received signals. The robustness of the proposed CS-2FFT-based algorithm in terms of the average positioning error is numerically evaluated using the realistic parameters in the IEEE 802.15.6-WBAN standard in the presence of additive white Gaussian noise. Due to the circular grid pattern and the CS-2FFT feature extraction method, the proposed scheme represents a significant reduction in the computational complexity, while improving the level of the resolution and the localization accuracy when compared to some classical CS-based positioning algorithms.

This paper proposes a novel method for extraction of discriminant spatio-spectral EEG features in motor imagery brain-computer interfaces. Considering a heteroscedastic binary classification setup, this method extracts the spatio-spectral features whose variance is maximized for one brain task and minimized for the other task. Therefore, our method can be considered as a spatio-spectral generalization of the conventional common spatial patterns (CSP) algorithm. In comparison to the similar solutions in the literature, such as filter-bank CSP (FBCSP) method, the proposed method benefits from joint processing of both spatial and spectral features, which improves the overall performance of the BCI while reducing its computational cost. Furthermore, our algorithm provides a simple measure that allows for ranking the discriminant power of extracted spatio-spectral features, which is not possible in FBCSP method. The experimental results demonstrate that the proposed method outperforms FBCSP for both raw EEG and preprocessed EEG data.

Linear discriminant analysis (LDA) is a commonly-used feature extraction technique. For matrix-variate data such as spatio-spectral electroencephalogram (EEG), matrix-variate LDA formulations have been proposed. Compared to the standard vector-variate LDA, these formulations assume a separable structure for the within-class and between-class scatter matrices; these structured parameters can be estimated more accurately with a limited number of training samples. However, separable scatters do not fit some data, resulting in aggravated performance for matrix-variate methods. This paper first proposes a common framework for the vector-variate LDA with non-separable scatters and our previously proposed solution with separable scatters. Then, a regularization of the non-separable scatter estimates toward the separable estimates is introduced. This novel regularized framework integrates vector-variate and matrix-variate approaches, and allows the estimated scatter matrices to adapt to the data characteristics. Experiments on data set V from BCI competition III demonstrate that the proposed framework achieves a considerable classification performance gain.

This paper investigates a novel combination of Co-occurrence of adjacent Local Binary Patterns histogram and Local Binary Patterns feature extraction methods for face detection in mobile phone applications. In particular, Co-occurrence of adjacent Local Binary Patterns histogram feature extraction provides exceptionally high discriminative power in face/non-face classification and hence is used to ensure the high accuracy of the proposed face detector. Local Binary Patterns feature extraction has low computation complexity and is thus used to reduce the overall processing speed. In the conducted face detection experiments, the proposed face detector yields comparable or better performance as well as faster computation speed than the existing best methods.

Recent findings in neuroscience have shown that the spectral components of electroencephalogram (EEG) signals convey information regarding the mental task not only in their power but also in their phase. This calls for the utilization of complex-valued spectrum, instead of the commonly used power spectral density, in designing the brain computer interfaces. This paper studies the complex-valued spectrum of the EEG signal recorded during mental imagery tasks, and provides a statistical model for the EEG spectral components. Motivated by the results of a recent work by the authors, this paper proposes a time-varying noncircularly-symmetric Gaussian model for complex-valued EEG spectrum during a mental imagery trial. It will be shown that the mean of this Gaussian model is constant over time, whereas its variance and pseudo-variance follow an autoregressive conditional heteroscedastic (ARCH) model. The validity of this model is then verified using statistical tests.

This paper introduces a novel eye region scrambling scheme capable of protecting privacy sensitive eye region information present in video contents. The proposed system consists of an automatic eye detection module followed by a privacy enabling JPEG XR encoder module. An object detection method based on a probabilistic model of image generation is used in conjunction with a skin-tone segmentation to accurately locate eye regions in real time. The utilized JPEG XR encoder effectively deteriorate the visual quality of privacy sensitive eye region at low computational cost. Performance of proposed solution is validated using benchmark face recognition algorithms on face image database. Experimental results indicate that the proposed solution is able to conceal identity by preventing successful identification at low computational costs.

Linear discriminant analysis (LDA), a conventional feature extraction technique, is a homoscedastic solution and ignores the second order information of the data. A heteroscedastic extension of LDA has been previously proposed which relies on the average pairwise Chernoff distances of the classes. However, in a multi-class scenario with number of classesC >; 2, the average of pairwise distances is not directly related to the classification error rate. Furthermore, the corresponding method imposes a high computational complexity of order O(C(C - 1)). This paper proposes an inherently multi-class heteroscedastic extension of LDA based on Matusita's separability measure, a multi-class generalization of the Chernoff distance which is related to multi-class error bounds. The proposed feature extractor can be trained non-iteratively with computational complexity of O(C). Experimental comparisons with the Chernoffmethod demonstrate both a performance improvement when estimated parameters are used, and a reduction of factor C - 1 in the computational load as predicted.

This work considers a CDMA-based Wireless Body Area Network (WBAN) where multiple biosensors communicate simultaneously to a central node in an asynchronous fashion. The main goal of this paper is to present an augmentation protocol for the physical layer of the IEEE 802.15.6 specifications with focus on the Multiple Access Interference (MAI) mitigation in a proactive WBAN. The proposed methodology uses a new set of orthogonal codes from the conventional Walsh-Hadamard matrix which has the special property of “cyclic orthogonality”. This property ensures that the asynchronous nature of the WBAN does not produce MAI amongst the multiple on-body sensors. The work investigates the optimality of such codes in WBANs from the link Bit Error Rate (BER) performance. We show that the proposed spreading codes outperform conventional non-cyclic orthogonal spreading codes in a practical Rayleigh fading environment.

This paper introduces a novel color filter array (CFA) image compression scheme capable of handling high dynamic range (HDR) representation. The proposed pipeline consists of a series of preprocessing steps followed by a JPEG XR encoding module. A 8-directional edge-sensing mechanism operator is used to reduce estimation error and preserve edge related information during color space conversion. The utilized YCoCg color space allows for a simplified pipeline implementation. Performance of the proposed solution is validated through objective quality assessment metrics and complexity analysis. Extensive experimentation reported in this paper indicates that the proposed solution is suitable for limited resource environments due to low complexity and high performance.

In this paper we propose a method to incorporate the inter-frame correlation to the problem of compressed video sensing (CVS) by means of high-order accuracy differential approximations. In particular, we encode high-frequency motion dynamics between consecutive frames in order to recover video data from under-sampled scheme. The proposed methodology couples the problem of compressed video sampling with motion dynamics to provide a unique protocol for CVS recovery. Numerical evaluations validate the proposed method by tracking high-frequency components between consecutive frames in video and effectively enhancing the recovery performance compared to the conventional CVS methods.

Emotion states greatly influence many areas in our daily lives, such as: learning, decision making and interaction with others. Therefore, the ability to detect and recognize one's emotional states is essential in intelligence Human Machine Interaction (HMI). The aim of this study was to develop a new system that can sense and communicate emotion changes expressed by the Central Nervous System (CNS) through the use of EEG signals. More specifically, this study was carried out to develop an EEG-based subject-dependent affect recognition system to quantitatively measure and categorize three affect states: Positively excited, neutral and negatively excited. In this paper, we discussed implementation issues associated with each key stage of a fully automated affect recognition system: emotion elicitation protocol, feature extraction and classification. EEG recordings from 5 subjects with IAPS images as stimuli from the eNTERFACE06 database were used for simulation purposes. Discriminating features were extracted in both time and frequency domains (statistical, narrow-band, HOC, and wavelet entropy) to better understand the oscillatory nature of the brain waves. Through the use of k Nearest Neighbor classifier (kNN), we obtained mean correct classification rates of 90.77% on the three emotion classes when K equals 5. This demonstrated the feasibility of brain waves as a mean to categorize a user's emotion state. Secondly, we also assessed the suitability of commercially available EEG headsets such as Emotive Epoc for emotion recognition applications. This study was carried out by comparing the sensor location, signal integrity with those of Biosemi Active II. A new set of recognition performance was presented with reduced number of channels.

In this study, the problem of transmitting an independent and identically distributed (i.i.d.) Gaussian source over an i.i.d. Gaussian wire-tap channel, with an i.i.d. Gaussian known interference available at the transmitter is considered. The intended receiver is assumed to have a certain minimum signal-to-noise ratio (SNR) and the eavesdropper is assumed to have a strictly lower SNR compared to the intended receiver. The objective is to minimise the distortion of source reconstruction at the intended receiver. In this study, an achievable distortion is derived when Shannon's source–channel separation coding scheme is used. Three hybrid digital–analogue secure joint source–channel coding schemes are then proposed, which achieve the same distortion. The first coding scheme is based on Costa's dirty-paper-coding scheme and wire-tap channel coding scheme, when the analogue source is not explicitly quantised. The second coding scheme is based on the superposition of the secure digital signal and the hybrid digital–analogue signal. It is shown that for the problem of communicating a Gaussian source over a Gaussian wire-tap channel with side information, there exists an infinite family of secure joint source–channel coding schemes. In the third coding scheme, the quantised signal and the analogue error signal are explicitly superimposed. It is shown that this scheme provides an infinite family of secure joint source–channel coding schemes with a variable number of binning. Finally, the proposed secure hybrid digital–analogue schemes are analysed under the main channel SNR mismatch. It is proven that the proposed schemes can give a graceful degradation of distortion with SNR under SNR mismatch, that is, when the actual SNR is larger than the designed SNR.

This letter provides a tractable bound for a perfect recovery condition in compressed sensing matrices using the spherical section property in the presence of side information. In particular, when the signal of interest is provided with side-information, we derive an equivalent semidefinite relaxation bound by introducing the related prior knowledge as an additional constraint to the semidefinite programming (SDP) problem. We recast a linear program (LP) cone to this problem and found the dual-SDP to be less complex to handle. Numerical evaluations on the proposed dual-SDP, validates the existence of sparse solutions with high- cardinalities.

This paper proposes new color local texture features, i.e., color local Gabor wavelets (CLGWs) and color local binary pattern (CLBP), for the purpose of face recognition (FR). The proposed color local texture features are able to exploit the discriminative information derived from spatiochromatic texture patterns of different spectral channels within a certain local face region. Furthermore, in order to maximize a complementary effect taken by using both color and texture information, the opponent color texture features that capture the texture patterns of spatial interactions between spectral channels are also incorporated into the generation of CLGW and CLBP. In addition, to perform the final classification, multiple color local texture features (each corresponding to the associated color band) are combined within a feature-level fusion framework. Extensive and comparative experiments have been conducted to evaluate our color local texture features for FR on five public face databases, i.e., CMU-PIE, Color FERET, XM2VTSDB, SCface, and FRGC 2.0. Experimental results show that FR approaches using color local texture features impressively yield better recognition rates than FR approaches using only color or texture information. Particularly, compared with grayscale texture features, the proposed color local texture features are able to provide excellent recognition rates for face images taken under severe variation in illumination, as well as for small- (low-) resolution face images. In addition, the feasibility of our color local texture features has been successfully demonstrated by making comparisons with other state-of-the-art color FR methods.

This paper proposes a novel face descriptor based on color information, i.e., so-called local color vector binary patterns (LCVBPs), for face recognition (FR). The proposed LCVBP consists of two discriminative patterns: color norm patterns and color angular patterns. In particular, we have designed a method for extracting color angular patterns, which enables to encode the discriminating texture patterns derived from spatial interactions among different spectral-band images. In order to perform FR tasks, the proposed LCVBP feature is generated by combining multiple features extracted from both color norm patterns and color angular patterns. Extensive and comparative experiments have been conducted to evaluate the proposed LCVBP feature on five public databases. Experimental results show that the proposed LCVBP feature is able to yield excellent FR performance for challenging face images. In addition, the effectiveness of the proposed LCVBP feature has successfully been tested by comparing other state-of-the-art face descriptors.

This paper proposes a new video face recognition (FR) method that is designed for significantly improving FR via adaptive fusion of multiple face features (belonging to the same subject) acquired from a face sequence of video frames. In this paper, we derive an upper bound for recognition error arising from the proposed weighted feature fusion to justify theoretically its effectiveness for recognition from videos. In addition, in order to compute the optimal weights of face features to be fused, we develop a novel weight determination solution based on fuzzy membership function and quality measurement for face images. Using four public video databases, the effectiveness of the proposed method has been successfully evaluated under the conditions that are similar to those in real-world video FR applications. Furthermore, our method is simple and straightforward to implement.

This paper presents a new technique of hidden Markov model (HMM) for video deinterlacing. Existing deinterlacing algorithms estimate missing pixels of an absent row in an interlaced frame on a sample-by-sample basis. In contrast, the proposed HMM-based deinterlacing technique adopts an approach of sequence estimation and makes a joint decision on the row of missing pixels as a whole. This allows a more thorough exploitation of the spatial correlation of the image signal. The HMM-based sequence estimation technique is coupled with a number of existing spatial deinterlacing algorithms in the literature to boost their performance. Experimental results show that HMM can significantly improve the deinterlacing results in both PSNR measure and subjective visual quality.

This paper proposes local color vector binary pattern (LCVBP) as a new color based face descriptor for face recognition (FR). In the proposed LCVBP, color pixel pattern of different spectral images is extracted as discriminative information. Further, to maximize a complementary effect taken from different color channels, color angle pattern between a pair of different spectral images is incorporated into the creation of LCVBP. Experimental results on three public databases (CMU-PIE, XM2VTSDB, and Color FERET) show that the proposed color texture feature is able to significantly improve FR performance, compared to the conventional grayscale texture features including Gabor wavelet and Local Binary Pattern (LBP). In particular, the proposed LCVBP is able to provide excellent FR performance for face images taken under severe illumination variation.

Over the years, a consensus has been reached that the introduction of 3D entertainment can only be a lasting success if the perceived image quality and the viewing comfort are better than those of conventional 2D television. There are different factors that affect the perceived quality of 3D content. In this paper, our objective is to obtain a good understanding of the effect that brightness has on the visual quality of 3D videos and compare it to that of the 2D. We capture outdoor and indoor scenes with different exposures and we perform subjective evaluation to investigate how brightness affects the perceived quality of the 3D experience.

The use of wireless body area networks requires correct delivery of the vital signs of the patient while managing precious energy in tiny biosensors. Toward these goals, we present an energy-efficient protocol suitable for the narrow band physical layer of the IEEE 802.15.6 standard in on-body sensor networks. This study considers a realistic channel model inspired by the Gilbert-Elliott channel including the erasure mode and a binary symmetric channel model. This work studies the feasibility of Raptor codes, the most efficient rateless codes, with the Frequency Shift Keying (FSK) modulation to overcome the reliability and power cost concerns in on-body sensor devices. The main advantage of using Raptor codes is to provide an inherent adaptive power management, due to the flexibility of the code rate and coding gain. Numerical results show that the Raptor coded FSK is more energy efficient and robust than that of the uncoded FSK and LDPC codes, in various channel realization, in particular, when patients make sequential position changes.

In this work, the problem of transmitting an i.i.d Gaussian source over an i.i.d Gaussian wiretap channel with an i.i.d Gaussian side information at the intended receiver is considered. The intended receiver is assumed to have a certain minimum SNR and the eavesdropper is assumed to have a strictly lower SNR compared to the intended receiver. The objective is minimizing the distortion of source reconstruction at the intended receiver. In this work, it is shown that the source-channel separation coding scheme is optimum in the sense of achieving the minimum distortion. A hybrid digital-analog Wyner-Ziv coding scheme is then proposed which achieve the minimum distortion. This secure joint source channel coding scheme is based on Wyner-Ziv coding scheme and wiretap channel coding scheme when the analog source is not explicitly quantized. The proposed secure hybrid digital-analog scheme is analyzed under the main channel SNR mismatch. It is proven that the proposed scheme can give a graceful degradation of distortion with SNR under SNR mismatch, i.e., when the actual SNR is larger than the designed SNR.

We consider a scenario where multiple collaborating cognitive radios (CR's) try to jointly detect spectrum opportunities in a wide-band spectrum within a predefined spectrum sensing time TS. Each CR is equipped with a tunable bandpass filter (BPF) and is able to sense one frequency band (channel) at a time. The sensing time consists of L sensing slots of length T. During each sensing slot, each of the collaborating CR's is assigned to sense one of the primary channels and report its observation to the other CR's. The goal is to maximize the expected number of identified idle channels by optimally choosing the channels to be sensed by each of the collaborating CR's at each sensing slot. We derive closed-form solutions for an optimal spectrum sensing policy and the associated reward function for the case of two collaborating CR's where individual sensing decisions are fused together according to the OR-rule. We show that the gain due to the optimal sensing policy is more significant when the spectrum utilization is high.

Owing to the unique characteristics of sensor devices, finding the energy-efficient modulation with a low-complexity implementation (refereed to as green modulation) poses significant challenges in the physical layer design of wireless sensor networks (WSNs). Towards this goal, the authors present an in-depth analysis on the energy efficiency of various modulation schemes using realistic models in the IEEE 802.15.4 standard to find the optimum distance-based scheme in a WSN over Rayleigh and Rician fading channels with path loss. The authors describe a proactive system model according to a flexible duty-cycling mechanism utilised in practical sensor apparatus. The present analysis includes the effect of the channel bandwidth and the active mode duration on the energy consumption of popular modulation designs. Path-loss exponent and DC–DC converter efficiency are also taken into consideration. In considering the energy efficiency and complexity, it is demonstrated that among various sinusoidal carrier-based modulations, the optimised non-coherent M-ary frequency shift keying (NC-MFSK) is the most energy-efficient scheme in sparse WSNs for each value of the path-loss exponent, where the optimisation is performed over the modulation parameters. In addition, the authors show that the on–off keying displays a significant energy saving as compared to the optimised NC-MFSK in dense WSNs with small values of path-loss exponent.

Adaptive Demodulation (ADM) is a new rate-adaptive system that operates without requiring Channel State Information (CSI) at the transmitter, instead using adaptive decision region boundaries at the receiver and encoding the data with a rateless code. This paper addresses the design and performance of an ADM scheme for two common differentially coherent schemes: M-DPSK and M-DAPSK. The optimal method for determining the most reliable bits for a given differential detection scheme is presented. In addition, simple (near-optimal) implementations are provided for recovering the most reliable bits from a received pair of differentially encoded symbols for systems using 16-DPSK and 16-DAPSK. The new receivers offer the advantages of a rate-adaptive system, without requiring CSI at the transmitter or a coherent phase reference at the receiver. Bit error analysis for the ADM system in both cases is presented along with numerical results of the spectral efficiency for the rate-adaptive systems operating over a Rayleigh fading channel.

This paper introduces the new color face recognition (FR) method that makes effective use of boosting learning as color-component feature selection framework. The proposed boosting color-component feature selection framework is designed for finding the best set of color-component features from various color spaces (or models), aiming to achieve the best FR performance for a given FR task. In addition, to facilitate the complementary effect of the selected color-component features for the purpose of color FR, they are combined using the proposed weighted feature fusion scheme. The effectiveness of our color FR method has been successfully evaluated on the following five public face databases (DBs): CMU-PIE, Color FERET, XM2VTSDB, SCface, and FRGC 2.0. Experimental results show that the results of the proposed method are impressively better than the results of other state-of-the-art color FR methods over different FR challenges including highly uncontrolled illumination, moderate pose variation, and small resolution face images.

The use of wireless implant technology requires correct delivery of the vital physiological signs of the patient along with the energy management in power-constrained devices. Toward these goals, we present an augmentation protocol for the physical layer of the medical implant communications service (MICS) with focus on the energy efficiency of deployed devices over the MICS frequency band. The present protocol uses the rateless code with the frequency-shift keying (FSK) modulation scheme to overcome the reliability and power cost concerns in tiny implantable sensors due to the considerable attenuation of propagated signals across the human body. In addition, the protocol allows a fast start-up time for the transceiver circuitry. The main advantage of using rateless codes is to provide an inherent adaptive duty cycling for power management, due to the flexibility of the rateless code rate. Analytical results demonstrate that an 80% energy saving is achievable with the proposed protocol when compared to the IEEE 802.15.4 physical layer standard with the same structure used for wireless sensor networks. Numerical results show that the optimized rateless coded FSK is more energy efficient than that of the uncoded FSK scheme for deep tissue (e.g., digestive endoscopy) applications, where the optimization is performed over modulation and coding parameters.

Using face annotation for effective management of personal photos in online social networks (OSNs) is currently of considerable practical interest. In this paper, we propose a novel collaborative face recognition (FR) framework, improving the accuracy of face annotation by effectively making use of multiple FR engines available in an OSN. Our collaborative FR framework consists of two major parts: selection of FR engines and merging (or fusion) of multiple FR results. The selection of FR engines aims at determining a set of personalized FR engines that are suitable for recognizing query face images belonging to a particular member of the OSN. For this purpose, we exploit both social network context in an OSN and social context in personal photo collections. In addition, to take advantage of the availability of multiple FR results retrieved from the selected FR engines, we devise two effective solutions for merging FR results, adopting traditional techniques for combining multiple classifier results. Experiments were conducted using 547 991 personal photos collected from an existing OSN. Our results demonstrate that the proposed collaborative FR method is able to significantly improve the accuracy of face annotation, compared to conventional FR approaches that only make use of a single FR engine. Further, we demonstrate that our collaborative FR framework has a low computational cost and comes with a design that is suited for deployment in a decentralized OSN.

A 3-stage algorithm is proposed for automatic detection of the four primary landmarks in 3D face imagery: eyes, nose, and mouth. Pose and facial expression variations which raise major difficulties in landmark processing are the primary focus of this work. In the first stage, Gaussian and Mean curvatures are used to extract ridge and valley points. The second stage utilizes a recursive grouping algorithm to generate candidate landmarks. In the last stage, a geometric model imposing a set of distance and angle constraints to the arrangement of candidate landmarks is utilized to select the final four landmarks. The algorithm is robust against variations in pose and expression with an overall success rate of 98.3%, using the Bosphorus Database as the test input.

This paper presents a novel approach for the visualization and clustering of crowd video contents by using multilinear principal component analysis (MPCA). In contrast to feature-point-based approach and frame-based dimensionality reduction approach, the proposed method maps each short video segment to a point in MPCA subspace to take temporal information into account naturally through tensorial representations. Specifically, MPCA projects each short segment of a video to a low-dimensional tensor first. A few MPCA features are then selected according to the variance captured as the final representation. Thus, a video is visualized as a trajectory in MPCA subspace. The trajectory generated enables visual interpretation of video content in a compact space as well as visual clustering of video events. The proposed method is evaluated on the PETS 2009 datasets through comparison with three existing methods for video visualization. The MPCA visualization shows superior performance in clustering segments of the same event as well as identifying the transitions between events.

In this paper, we propose a new color face recognition (FR) method which effectively employs feature selection algorithm in order to find the set of optimal color components (from various color models) for FR purpose. The proposed FR method is also designed to improve FR accuracy by combining the selected color components at the feature level. The effectiveness of the proposed color FR method has been successfully demonstrated using two public CMU-PIE and Color FERET face databases (DB). In our comparative experiments, traditional grayscale-based FR, previous color-based FR, and popular local binary pattern (LBP) based FR methods were compared with the proposed method. Experimental results show that our color FR method performs better than the aforementioned three different FR approaches. In particular, the proposed method can achieve 7.81% and 18.57% improvement in FR performance on the CMU-PIE and Color FERET DB, respectively, compared to representative color-based FR solutions previously developed.

Face detection is becoming popular in surveillance applications; however, the need of enormous size face/non-face dataset, large number of features, and long training time are persistent problems. This paper claims that only a subset of the total number of features conserves the major power to detect faces; hence, this subset is capable to detect faces with high detection rate. The proposed detector fuses the results of two classifiers where one is trained with only 40 Haar-like features and the other is trained with only 50 LBP Histogram features. A pre-processing stage of skin-tone detection is applied to reduce the false positive rate. The detector is examined on real-life low-resolution surveillance sequence. Conducted experiments show that the proposed detector can achieve a high detection rate and a low false positive rate. Also, it outperforms Lienhart detector and tolerates wide range of illumination and blurring changes.

Due to unique characteristics of sensor nodes, choosing an energy-efficient modulation scheme with low-complexity implementation (refereed to as green modulation) is a critical factor in the physical layer of Wireless Sensor Networks (WSNs). The main goal of this paper is to analyze and compare the energy efficiency of various sinusoidal carrier-based modulation schemes using parameters in the IEEE 802.15.4 standard and state-of-the art technology to find the best scheme in a dense WSN over frequency-flat Rayleigh fading channel with path-loss. Experimental results show that M-ary Frequency Shift Keying (MFSK) with small order of M has significant energy saving compared to OQPSK and MQAM for short range scenarios, and could be considered as a realistic candidate in dense WSNs. In addition, MFSK has the advantage of less complexity and cost in implementation than the other schemes.

In this work, we propose a new and novel framework for improving the performance of linear feature extraction (LFE) algorithms, characterized by the Bayesian error probability (BEP) in the extracted feature domain. The proposed framework relies on optimizing a tight quadratic approximation to the BEP in the transformed space with respect to the transformation matrix. Applied to many synthetic multi-class Gaussian classification problems, the proposed optimization procedure significantly improves the classification performance when it is initialized by popular LFE matrices such as the Fisher linear discriminant analysis.

Cognitive radios try to exploit “blank spaces” in the licensed band which are not being used by primary users at a particular place and time. In the absence of cooperation between the primary and the secondary networks, spectrum sensing enables secondary users to monitor a licensed band in order to find idle channels for opportunistic access. In this work, we propose a two-stage spectrum detection strategy that decreases the average channel search time by allowing the spectrum detector to focus on frequency channels which are more likely to be vacant. We show that the proposed detection strategy significantly outperforms the conventional single-stage strategy when the spectrum utilization is high.

This paper investigates the inconvenience of using huge number of features, enormous training dataset and lengthy training session to achieve a good performance frontal face detector. The proposed face detector is based on a novel idea which proposes using joint decision from two parallel different features trained detectors, one detector is trained with Local Binary Patterns (LBP) features and the other with Haar-like features. Both detectors are trained with few features using not a huge face/non-face dataset and within relatively short period of time. Hence, both detectors agree on the face image but seldom agree on the non-face image. The result is significantly improved using a multi-detections merging algorithm using simple clustering method. The robustness of the detector is examined once using a face/non-face dataset and compared to Lienhart frontal face detector, and secondly using a real-life sequence.

The objective in feature extraction is to compress the data while maintaining the same Bayes classification error as on the original data. This objective is achieved by a sufficient statistic with the minimum dimension. This paper derives a non-iterative linear feature extractor that approximates the minimal-dimension linear sufficient statistic operator for the classification of Gaussian distributions. This new framework alleviates the bias of an existing similar formulation towards the parameters of a reference class. Moreover, it is a heteroscedastic extension of linear discriminant analysis and captures the discriminative information in the first and second central moments of the data. The proposed method can improve the performance of the similar feature extractors while imposing equal, or even lower, computational complexity.

This paper proposes a directional image interpolation technique based on analysis of variance (ANOVA). ANOVA, a hypothesis testing methodology, is used to decide on the interpolation direction. Experimental results show the proposed ANOVA-based image interpolation technique preserves edges and fine structures of an image and suppresses common interpolation artifacts (e.g. ringing, blurring, and jaggies).

In this paper, we propose a novel feature representation based on color-based Local Binary Pattern (LBP) texture analysis for face recognition (FR). The proposed method exploits both color and texture discriminative features of a face image for FR purpose. We evaluate the proposed feature using three public face databases: CMU-PIE, Color FERET, and XM2VTSDB. Experimental results show that the results of the proposed feature impressively better than the results of grayscale LBP and color features. In particular, it is shown that the proposed feature is highly robust against severe variations in illumination and spatial resolution.

Face detection in video sequence is becoming popular in surveillance applications, but the usage of large number of features and the long training time are persistent problems. This paper integrates two types of Local Binary Patterns (LBP) features in order to achieve a high detection rate with a high discriminative power face detector. First LBP feature is a novel way of using the Circular LBP, in which the pixels of the image are targeted; it is a non-computationally expensive feature extraction. The second LBP feature is the LBP Histogram, in which regions in the image are targeted; it is more computationally expensive than Circular LBP features but has higher discriminative power. The proposed detector is examined on real-life low-resolution surveillance sequence. Conducted experiments show that the proposed detector achieves 98% detection rate in comparison to 91% for the Lienhart detector. The proposed detector tolerates wide range of illumination changes.

This paper proposes a novel weighted feature fusion in color face recognition (FR) to automatically annotate faces in personal videos. In the proposed FR method, multiple face images (belonging to the same subject) are clustered from a sequence of video frames. To facilitate a complementary effect on improving annotation performance, the grouped faces are combined using the proposed weighted feature fusion. In addition, we make effective use of facial color feature to cope with decrease in annotation performance due to a low-resolution face in personal videos. To evaluate the effectiveness of proposed FR method, more than 40,000 video frames for 10 real-world personal videos are collected from an existing online video sharing website. Experimental results show that the proposed FR method significantly improves annotation performance obtained using conventional grayscale image based FR methods.

Consider a homoscedastic multi-class Gaussian classification problem where the class mean vectors and the common covariance matrix are not known to the practitioner. Rather, they are estimated from given sample vectors available for each class. In this paper, an empirical procedure for approximating the bias of the holdout estimator of the Bayesian error probability (BEP) is presented. Synthetic experiments demonstrate the accuracy of the proposed procedure and how it can be used for guiding the practitioner about the necessary amount of data vectors required to achieve a certain level of accuracy in the BEP estimation. When applied to real world classification problems from the UCI machine learning repository, the proposed procedure was successfully used to estimate the test error probability based on the training data only. Moreover, with a reasonable degree of accuracy, the proposed procedure predicted the test BEP when the amount of the training data in increased.

Classification of mental tasks from electroencephalogram (EEG) signals has important applications in brain-computer interfacing (BCI). However, classification of the highly redundant and high-dimensional EEG signal, with high spatial and spectral correlations, is quite challenging. Therefore, the discriminant information, especially that of the first and second data moments, need to be extracted in the form of uncorrelated features. This work addresses this need by approximating a linear minimal-dimension sufficient statistic of the EEG matrix data in both spatial and spectral domains. As a result of the two-dimensional spatio-temporal approach and the generalized sufficiency approximation, a significant improvement on the classification accuracy is achieved.

A local algorithm is proposed and analyzed to monitor the health of a wireless sensor network. Our previous algorithm, the average consensus-based algorithm, works based on the mean estimator. Therefore, it fails to detect the defective nodes in the presence of large outliers. However, the algorithm proposed in this paper works based on the median estimator, and is able to detect the defective nodes, even in the presence of a large number of outliers. To calculate the median in a distributed scheme, a nonlinear consensus algorithm is proposed. By applying the nonlinear consensus algorithm, all the nodes in the network compute the median iteratively, using only local communications. We show that the proposed algorithm is more robust than the previous algorithms in this area and outperforms all of them. The simulation results verify the effectiveness of the proposed algorithm in calculating the median in a distributed scheme and in detecting the defective nodes in the network.

In this paper, a novel face annotation framework is proposed that systematically leverages context information such as situation awareness information with current face recognition (FR) solutions. In particular, unsupervised situation and subject clustering techniques have been developed that are aided by context information. Situation clustering groups together photos that are similar in terms of capture time and visual content, allowing for the reliable use of visual context information during subject clustering. The aim of subject clustering is to merge multiple face images that belong to the same individual. To take advantage of the availability of multiple face images for a particular individual, we propose effective FR methods that are based on face information fusion strategies. The performance of the proposed annotation method has been evaluated using a variety of photo sets. The photo sets were constructed using 1385 photos from the MPEG-7 Visual Core Experiment 3 (VCE-3) data set and approximately 20000 photos collected from well-known photo-sharing websites. The reported experimental results show that the proposed face annotation method significantly outperforms traditional face annotation solutions at no additional computational cost, with accuracy gains of up to 25% for particular cases.

Biometric encryption (BE) has recently been identified as a promising paradigm to deliver security and privacy, with unique technical merits and encouraging social implications. An integral component in BE is a key binding method, which is the process of securely combining a signal, containing sensitive information to be protected (i.e., the key), with another signal derived from physiological features (i.e., the biometric). A challenge to this approach is the high degree of noise and variability present in physiological signals. As such, fuzzy methods are needed to enable proper operations, with adequate performance results in terms of false acceptance rate and false rejection rate. In this work, the focus will be on a class of fuzzy key binding methods based on dirty paper coding known as quantization index modulation. While the methods presented are applicable to a wide range of biometric modalities, the face biometric is selected for illustrative purposes, in evaluating the QIM-based solutions for BE systems. Performance evaluation of the investigated methods is reported using data from the CMU PIE face database.

Indoor positioning is an enabling technology for delivery of location-based services in mobile computing environments. This paper proposes a positioning solution using received signal strength in indoor wireless local area networks. In this application, an explicit measurement equation and the corresponding noise statistics are unknown because of the complexity of the indoor propagation channel. To address these challenges, we introduce a new state-space Bayesian filter: the nonparametric information (NI) filter. This filter effectively tracks motion in situations where the Kalman filter and its variants are inapplicable, while maintaining a computational complexity comparable to that of the Kalman filter. To deal with the noisy nature of the indoor propagation environment, the NI filter is used in the design of an intelligent dynamic WLAN tracking system. The system anticipates future position values and adapts its sensing and estimation parameters accordingly. Our experimental results conducted on measurements from a real office environment indicate that the combination of the intelligent design and the NI filter results in significant improvements over the Kalman and particle filters.

Changeability and privacy protection are important factors for widespread deployment of biometrics-based verification systems. This paper presents a systematic analysis of a random-projection (RP)-based method for addressing these problems. The employed method transforms biometric data using a random matrix with each entry an independent and identically distributed Gaussian random variable. The similarity- and privacy-preserving properties, as well as the changeability of the biometric information in the transformed domain, are analyzed in detail. Specifically, RP on both high-dimensional image vectors and dimensionality-reduced feature vectors is discussed and compared. A vector translation method is proposed to improve the changeability of the generated templates. The feasibility of the introduced solution is well supported by detailed theoretical analyses. Extensive experimentation on a face-based biometric verification problem shows the effectiveness of the proposed method.

This paper proposes a new receiver structure for linear-dispersion (LD) codes, subsuming orthogonal, quasiorthogonal and V-BLAST codes. We suggest to use widely-linear minimum-mean-squared-error (WL-MMSE) estimates of transmitted symbols in lieu of the sufficient statistics for maximum likelihood (ML) detection of these symbols. Proposed structure offers both optimal (ML) and suboptimal solutions. Simulation results show that the suboptimal receiver performs close to the optimal one, while reducing the receiver's complexity. Structure of the proposed receiver is particularly studied for orthogonal and quasi-orthogonal LD codes. Specifically, it is proved that Alamouti's combining scheme provides WL-MMSE estimates of the transmitted symbols.

In this letter, a design metric of the interleaver is derived from the union bound of the bit error rate (BER) of linearly precoded orthogonal frequency division multiplexing (OFDM) systems, assuming that the power-delay profile of the channel is known at the transmitter. A new cyclic shift interleaving scheme is proposed to improve the BER performance of OFDM systems. The proposed interleaving scheme is parameterized, and it incorporates existing interleaving schemes. The simulation results indicate that the proposed interleavers outperform the existing interleavers by up to 1.5 dB in a practical scenario.

Thanks to high-speed Internet access and feature-rich mobile devices, the demand for ubiquitous and secure surveillance systems has increased. In this paper, we propose a privacy-protected video surveillance system that makes use of scalable video coding (SVC). SVC can be used to fulfill the requirement of omnipresence. Further, to address privacy concerns, we detect face regions and subsequently scramble these regions-of-interest (ROIs) in the compressed domain. To demonstrate the feasibility of the proposed video surveillance system, simulation results are provided. The results show that our system is able to provide a good level of security, while offering access to surveillance video content in heterogeneous usage environments.

An important parameter in analysis of physiological tremor is the diagnosis and study of neurological disorders. The instantaneous tremor frequency (ITF) is an important parameter in tremor analysis. This paper proposes a novel stochastic filter, the multiple extended Kalman filter (M-EKF), for tracking of ITF from neural microelectrode recordings. The M-EKF mitigates degradations in filter performance resulting from a mismatch between assumed initial conditions and those of a particular realization of a stochastic system. Specifically, the M-EKF is comprised of a bank of extended Kalman filters (EKF), each initialized with different conditions, selected according to the unscented transform. The final estimate is a weighted average of the individual estimates provided by each EKF where the weights reflect how closely the assumed EKF initial conditions match those of the true system. The M-EKF is applied to a synthetic tremor model to display its superior performance to that of the EKF and the unscented Kalman filter.

Face recognition (FR) for emerging applications such as face tagging for social networking, consumer products, and gamming utilize color images stored in distributed repositories. Such images are often in compressed format and of different dimensions. This compression mismatch problem may adversely affect the performance of the face recognition engine. In this paper, we present a comparative investigation of the image compression mismatch problem. Two commonly used color image based face recognition solutions are utilized. More than three thousand images of 341 subjects, typical of the problem, are collected from three public databases. The experimental results support the main thesis of the paper that recognition performance depends critically on the color image properties.

Fourier Coefficients have long been used to achieve invariance to signal transformations. For the purposes of image processing, the magnitude of the Fourier transform has been used in conjunction with other transforms to achieve invariance to rotation. In this paper we propose a Rotation Invariant Descriptor for matching images based on features derived from the Discrete Fourier Transform (DFT). The features combine both the phase and the magnitude information to achieve invariance. Experiments are conducted to show the robustness of these features under changes of scale and compression of images.

The increasing popularity of user-generated content (UGC) requires effective annotation techniques in order to facilitate precise content search and retrieval. In this paper, we propose a new approach for the semantic annotation of personal video content, taking advantage of user-contributed tags available in an image folksonomy. Video shots and folksonomy images are first represented by a semantic vector. Next, the semantic vectors are used to measure the semantic similarity between each video shot and the folksonomy images. Tags assigned to semantically similar folksonomy images are then used to annotate the video shots. To verify the effectiveness of the proposed annotation method, experiments were performed with video sequences retrieved from YouTube and images downloaded from Flickr. Our experimental results demonstrate that the proposed method is able to successfully annotate personal video content with user-contributed tags retrieved from an image folksonomy. In addition, the size of our tag vocabulary is significantly higher than the size of the tag vocabulary used by conventional annotation methods.

This paper proposes an uncorrelated multilinear discriminant analysis (UMLDA) framework for the recognition of multidimensional objects, known as tensor objects. Uncorrelated features are desirable in recognition tasks since they contain minimum redundancy and ensure independence of features. The UMLDA aims to extract uncorrelated discriminative features directly from tensorial data through solving a tensor-to-vector projection. The solution consists of sequential iterative processes based on the alternating projection method, and an adaptive regularization procedure is incorporated to enhance the performance in the small sample size (SSS) scenario. A simple nearest-neighbor classifier is employed for classification. Furthermore, exploiting the complementary information from differently initialized and regularized UMLDA recognizers, an aggregation scheme is adopted to combine them at the matching score level, resulting in enhanced generalization performance while alleviating the regularization parameter selection problem. The UMLDA-based recognition algorithm is then empirically shown on face and gait recognition tasks to outperform four multilinear subspace solutions (MPCA, DATER, GTDA, TR1DA) and four linear subspace solutions (Bayesian, LDA, ULDA, R-JD-LDA).

This paper proposes an uncorrelated multilinear principal component analysis (UMPCA) algorithm for unsupervised subspace learning of tensorial data. It should be viewed as a multilinear extension of the classical principal component analysis (PCA) framework. Through successive variance maximization, UMPCA seeks a tensor-to-vector projection (TVP) that captures most of the variation in the original tensorial input while producing uncorrelated features. The solution consists of sequential iterative steps based on the alternating projection method. In addition to deriving the UMPCA framework, this work offers a way to systematically determine the maximum number of uncorrelated multilinear features that can be extracted by the method. UMPCA is compared against the baseline PCA solution and its five state-of-the-art multilinear extensions, namely two-dimensional PCA (2DPCA), concurrent subspaces analysis (CSA), tensor rank-one decomposition (TROD), generalized PCA (GPCA), and multilinear PCA (MPCA), on the tasks of unsupervised face and gait recognition. Experimental results included in this paper suggest that UMPCA is particularly effective in determining the low-dimensional projection space needed in such recognition tasks.

In many current face-recognition (FR) applications, such as video surveillance security and content annotation in a web environment, low-resolution faces are commonly encountered and negatively impact on reliable recognition performance. In particular, the recognition accuracy of current intensity-based FR systems can significantly drop off if the resolution of facial images is smaller than a certain level (e.g., less than 20 x 20 pixels). To cope with low-resolution faces, we demonstrate that facial color cue can significantly improve recognition performance compared with intensity-based features. The contribution of this paper is twofold. First, a new metric called "variation ratio gain" (VRG) is proposed to prove theoretically the significance of color effect on low-resolution faces within well-known subspace FR frameworks; VRG quantitatively characterizes how color features affect the recognition performance with respect to changes in face resolution. Second, we conduct extensive performance evaluation studies to show the effectiveness of color on low-resolution faces. In particular, more than 3000 color facial images of 341 subjects, which are collected from three standard face databases, are used to perform the comparative studies of color effect on face resolutions to be possibly confronted in real-world FR systems. The effectiveness of color on low-resolution faces has successfully been tested on three representative subspace FR methods, including the eigenfaces, the fisherfaces, and the Bayesian. Experimental results show that color features decrease the recognition error rate by at least an order of magnitude over intensity-driven features when low-resolution faces (25 x 25 pixels or less) are applied to three FR methods.

In video-based face recognition (FR) applications such as surveillance security, the resolution of a facial image could significantly impact the reliability of recognition system. In most practical FR applications, it is reasonable to assume that the face resolution used during training is higher than the resolution used during the identification or verification process. This dimensional mismatch negatively impacts the performance of the traditional subspace recognition solutions. To address resolution mismatch problem, this paper introduces a novel estimation method capable of determining feature subspace of the lower resolution probe given an eigenspace pre-trained with higher resolution facial images. The effectiveness of the proposed solution has been successfully tested on standard CMU PIE face dataset. Experimental results and comparative evaluations provided in this work demonstrate the benefits of the proposed solution.

In a wide range of communication systems, including DS-CDMA and OFDM systems, the signal-of-interest might be corrupted by an improper (F.D. Neeser et al.,1993) (also called non circularly symmetric (B. Picinbono, 1994)) interfering signal. This paper studies the maximum likelihood (ML) detection of binary signals in the presence of additive improper complex Gaussian noise. Proposing a new measure for noncircularity of complex random variables, we will derive the ML decision rule and its performance based on this measure. It will be shown that the ML detector performs pseudo correlation (F.D. Neeser et al.,1993) as well as conventional correlation of the observation to the signals-of-interest. As an alternative solution, we will propose a filter for converting improper signals to proper ones, called circularization filter, and will utilize it together with a conventional matched-filter (MF) to construct an ML detector.

This paper introduces a novel, cost effective visual cryptography scheme suitable for color image transmission over bandwidth constraint channels. Unlike previously proposed schemes, the solution offers perfect reconstruction while producing shares with size smaller than that of the input image. The maximum distance separable (MDS) code principle used in the design allows for the introduction of a flexible framework that compares favorably to competing solutions as it can be seen by examining the experimental results included in this paper.

Positioning in wireless networks has gained significant ground as an enabling technology for various applications such as event detection and context awareness. Since these positioning systems rely on radio features to locate a mobile, they are susceptible to non-cryptographic attacks resulting from malicious alteration of the propagation environment. This paper proposes a sensor selection scheme for increasing the resilience of fingerprinting-based positioning systems to RSS-based attacks in the context of wireless local area networks (WLAN). A distributed positioning scheme is proposed whereby an estimate is obtained from each WLAN access point (AP). Sensor selection is performed based on a nonparametric estimate of the Fisher information. Experimental results indicate superior performance compared to existing methods and graceful performance degradation in presence of RSS attacks.

A novel, cost effective visual secret sharing (VSS) scheme is introduced in this work. The scheme partitions the input image into sub-components by utilizing a segmentation step. The contents of the segmented parts are subsequently used to form a secret sharing solution. This modular approach, which mimics the second generation coding principles, affords perfect reconstruction. Comparisons and experimental results included here indicate that it also compares very favorably against competing solutions when it comes to pixel expansion.

This paper presents a scheme for secure coding of arbitrarily-shaped visual objects. Called SecST-SPIHT, it employs SPIHT based coding along with selective encryption for efficient, secure storage and transmission of visual object shape and texture. The selective encryption utilizes a novel bit classification scheme which ensures protection of the entire code by encrypting only a small number of bits (less than 5% for securing shape and texture, and less than 0.5% for just the texture). The encryption is performed in the compressed domain and does not affect the rate-distortion performance of the coder. A parameter allows control over the strength of the encryption versus required processing overhead. The scheme can be employed in a privacy protected surveillance system, whereby visual objects of human subjects are encrypted so that the content is only available to certain entities, such as persons of authority, possessing the correct decryption key.

Tensorial data are frequently encountered in various machine learning tasks today and dimensionality reduction is one of their most important applications. This paper extends the classical principal component analysis (PCA) to its multilinear version by proposing a novel unsupervised dimensionality reduction algorithm for tensorial data, named as uncorrelated multilinear PCA (UMPCA). UMPCA seeks a tensor-to-vector projection that captures most of the variation in the original tensorial input while producing uncorrelated features through successive variance maximization. We evaluate the UMPCA on a second-order tensorial problem, face recognition, and the experimental results show its superiority, especially in low-dimensional spaces, through the comparison with three other PCA-based algorithms.

In the practical face recognition (FR) applications, low-resolution faces (20 times 20 pixels or less) are commonly encountered and negatively impact on reliable performance. To overcome low-resolution face problem, we show that face color can significantly improve the performance compared to intensity-based features. The contribution of this paper is twofold. First, a new metric called dasiavariation ratio gainpsila (VRG) is proposed to theoretically prove the significance of color effect on low-resolution faces. Second, we conduct extensive performance comparison studies. In particular, 3,192 color facial images corresponding to 341 subjects, collected from three standard CMU PIE, FERET, and XM2VTSDB face databases, were used to perform comparative studies of color effect on various face resolutions. Experimental results verified that face color feature improves the degraded recognition rate due to low-resolution faces by at least an order of magnitude over intensity-based features.

Most consumer digital cameras use a single light sensor which captures color information using a color filter array (CFA). This produces a mosaic image, where each pixel location contains a sample of only one of three colors, either red, green or blue. The two missing colors at each pixel location must be interpolated from the surrounding samples in a process called demosaicking. The conventional approach to compressing video captured with these devices is to first perform demosaicking and then compress the resulting full-color video using standard methods. In this paper two methods for compressing CFA video prior to demosaicking are proposed. In our first method, the CFA video is directly compressed with the H.264 video coding standard in 4:2:2 sampling mode. Our second method uses a modified version of H.264, where motion compensation is altered to take advantage of the properties of CFA data. Simulations show both proposed methods give better compression efficiency than the demosaick-first approach at high bit rates, and thus are suitable for applications, such as digital camcorders, where high quality video is required.

This paper introduces a multilinear principal component analysis (MPCA) framework for tensor object feature extraction. Objects of interest in many computer vision and pattern recognition applications, such as 2-D/3-D images and video sequences are naturally described as tensors or multilinear arrays. The proposed framework performs feature extraction by determining a multilinear projection that captures most of the original tensorial input variation. The solution is iterative in nature and it proceeds by decomposing the original problem to a series of multiple projection subproblems. As part of this work, methods for subspace dimensionality determination are proposed and analyzed. It is shown that the MPCA framework discussed in this work supplants existing heterogeneous solutions such as the classical principal component analysis (PCA) and its 2-D variant (2-D PCA). Finally, a tensor object recognition system is proposed with the introduction of a discriminative tensor feature selection mechanism and a novel classification strategy, and applied to the problem of gait recognition. Results presented here indicate MPCA's utility as a feature extraction tool. It is shown that even without a fully optimized design, an MPCA-based gait recognition module achieves highly competitive performance and compares favorably to the state-of-the-art gait recognizers.

Adaptive demodulation (ADM) is a new rate adaptive technique wherein the receiver demodulates only the most reliable bits and treats unreliable bits as erasures. This paper derives the optimum and simple near-optimum receivers for an ADM system operating without a coherent phase reference, where differential encoding is assumed at the transmitter (using 16-DPSK and 16-DAPSK). The new receivers offer the advantages of a rate-adaptive system, without requiring channel state information at the transmitter or a coherent phase reference at the receiver. Bit error analysis for the ADM system in both cases is presented along with numerical results of the spectral efficiency for the rate adaptive systems operating over a Rayleigh fading channel.

This paper introduces a precoder design for wireless receivers with imperfect fading channel state information (CSI). Imperfect CSI arises from inaccurate channel modelling or suboptimum channel estimation procedure. Precoders at transmitters are considered here as the means to reduce detection error. To that end, a new lower bound for pairwise error probability (PEP) is introduced. Using the proposed lower bound, precoders are optimized to improve the bit error rate (BER) performance of a receiver which uses basis expansion models (BEM) to estimate time varying channel coefficients without statistical information about the channel. Two precoder design criteria, namely pairwise coding and diversity gain are introduced through the proposed lower bound. Experimentation and theoretical analysis reported here indicate significant improvement in the BER performance of the BEM based receiver when it is used with the precoder in fading channels.

The recent proliferation of Location-Based Services (LBSs) has necessitated the development of effective indoor positioning solutions. In such a context, Wireless Local Area Network (WLAN) positioning is a particularly viable solution in terms of hardware and installation costs due to the ubiquity of WLAN infrastructures. This paper examines three aspects of the problem of indoor WLAN positioning using received signal strength (RSS). First, we show that, due to the variability of RSS features over space, a spatially localized positioning method leads to improved positioning results. Second, we explore the problem of access point (AP) selection for positioning and demonstrate the need for further research in this area. Third, we present a kernelized distance calculation algorithm for comparing RSS observations to RSS training records. Experimental results indicate that the proposed system leads to a 17 percent (0.56 m) improvement over the widely used K-nearest neighbor and histogram-based methods.

This paper proposes a dynamic MMSE estimator for tracking mobile users in indoor wireless local area networks (WLAN) based on received signal strength (RSS). The method uses a training-based static estimate obtained by an adaptive kernel density estimator as the input into a Kalman Filter. Predictions from the filter are used during the next iteration to adaptively select a subset of training data, contained in a radio map, for the static estimator. Experimental results show that the combination of the Kalman filter and the adaptive radio map technique results in nearly 0.5 m (20%) improvement in root mean square location accuracy when compared to static localization

This paper presents a spatiotemporal demosaicking scheme suitable for digital video cameras. The proposed method uses multi-stage processing concepts to follow varying spatiotemporal characteristics of the captured video. The method also uses both the structural and spectral information during demosaicking in order to produce visually pleasing full-color videos

In this paper, we present several algorithms developed for restoration of motion blurred images. We begin with a single-image based deblurring approach in the case of linear constant motion. This approach is a wavelet-based method with a novel lp-norm regularization term. Due to the introduction of the wavelet and regularization techniques, the approach is rather robust against noise amplification during deconvolution. Then, we further developed a general multi-image based deblurring framework improved from recent works of Rav-Acha, A et al., (2005). The proposed framework is able to effectively take advantage of information contained in the multiple input images, even when they are blurred in the same direction-a case hard to be dealt with by traditional solutions. The proposed methods are evaluated on both simulated and real data, and the obtained experimental results indicate promising results

This paper presents a localized coarse-to-fine algorithm for efficient and accurate pedestrian localization and silhouette extraction for the gait challenge data sets. The coarse detection phase is simple and fast. It locates the target quickly based on temporal differences and some knowledge on the human target. Based on this coarse detection, the fine detection phase applies a robust background subtraction algorithm to the coarse target regions and the detection obtained is further processed to produce the final results. This algorithm has been tested on 285 outdoor sequences from the gait challenge data sets, with wide variety of capture conditions. The pedestrian targets are localized very well and silhouettes extracted resemble the manually labeled silhouettes closely

This paper presents a cost-effective image sharpening solution suitable for single-sensor digital cameras. The proposed solution enhances the structural content of the sensor image data captured using a Bayer color filter array (CFA). Subsequent demosaicking of the enhanced CFA image produces a visually-pleasing demosaicked image without the need for expensive sharpening in the RGB color domain. Simulation studies indicate that the proposed imaging pipeline, which employs an image sharpening step before the actual demosaikcking module, yields excellent performance and outperforms the conventional pipeline (demosaicking followed by full-color image sharpening) in terms of both subjective and objective image quality measures.

A new edge-sensing demosaicking solution based on a novel vector spectral model is introduced in this paper. Using the vector spectral model, the proposed solution preserves the magnitude and the directional characteristics of the single-sensor captured image in both smooth and edge areas. Experimentation reported in this paper indicates that the proposed demosaicking solution produces visually pleasing color images and can outperform the powerful demosaicking schemes in terms of the performance.

It is well-known that supervised learning techniques such as linear discriminant analysis (LDA) often suffer from the so called small sample size problem when apply to solve face recognition problems. This is due to the fact that in most cases, the number of training samples is much smaller than the dimensionality of the sample space. The problem becomes even more severe if only one training sample is available for each subject. In this paper, followed by the well-known unsupervised technique, kernel principal component analysis (KPCA), a novel feature selection scheme is proposed to establish a discriminant feature subspace in which the class separability is maximized. Extensive experiments performed on the FERET database indicate that the proposed scheme significantly boosts the recognition performance of the traditional KPCA solution

This paper presents a powerful framework for cDNA microarray spot localization. Following the multichannel nature of the cDNA image data, the framework allows the utilization of various scalar and vector edge operators to determine edges, discontinuities and structural elements, and generate edge maps of cDNA microarray images. It will be shown that vector edge operators, which utilize both the spectral and spatial characteristics of cDNA image, outperform well-known scalar edge detectors. In addition, built on robust theory order statistics, vector operators are more immune to the noise present in microarray images than common scalar operators.

We introduce a rate-adaptive system in which the receiver demodulates only those bits that have a high probability of being correct, treating nondemodulated bits as erasures. Several sets of decision regions, derived using composite hypothesis testing, are proposed for 16-QAM and 16-phase-shift keying, which allow for the simple implementation of this demodulation strategy. We demonstrate that pre-encoding the data with a Raptor code allows for simple reconstruction of the message, regardless of the erasure pattern introduced from the nondemodulated bits. We prove the optimality of the proposed decision regions in selecting the most likely subset of bits from any received symbol in moderate-to-high signal-to-noise ratios, and we analyze the performance of demodulating with these decision regions over an additive white Gaussian noise channel. Also demonstrated is the strong performance of 16-QAM for this application, compared with other power-efficient constellations and the near-optimality of using Gray mapping, even under the proposed alternate sets of decision regions.

The Bell Labs layered space-time (BLAST) architecture is a simple and efficient multiantenna coding structure that can achieve high spectral efficiency. Many BLAST detectors require more receiver antennas than transmitter antennas. We propose two novel turbo-processing BLAST detectors that can operate in systems with fewer receiver antennas than transmitter antennas. Both detectors are based on the group-detection strategy. The first proposed detector, the reduced-dimension maximum a posteriori (RDMAP) detector uses a dynamically formed group for each bit decision, while the second proposed detector, the group maximum a posteriori (GMAP) uses a static grouping. For both detectors, a maximum a posteriori (MAP) decision is made using a group of transmitted symbols, and the remaining signal contribution is treated as interference. The interference is characterized as nonzero mean colored-noise source that is whitened before a decision is made. Both proposed detectors are generalizations of the MAP detector and the turbo-processing minimum mean-squared error (MMSE) detector in Sellathurai and Haykin, and Abe and Matsumoto. An uncoded bit-error rate analysis for an independent Rayleigh fading environment is also presented. Simulated results are presented which show that both the RDMAP and GMAP detectors have a performance improvement over the MMSE detector, especially in systems having an excess number of transmitter antennas.

Pulse shaping is examined as a means to improve the performance of a differential offset quadrature phase-shift keying system in a bandwidth-constrained environment. Through optimization with respect to a composite Nyquist criterion, the derived pulse shapes have comparable performance to a /spl pi//4-differential quadrature phase-shift keying in an additive white Gaussian noise (AWGN) channel and better performance in a hard-limited AWGN channel.

This paper studies an enhanced widely linear (WL) receiver for direct-sequence code-division multiple-access (DS-CDMA) systems that employ aperiodic spreading sequences with offset quadrature phase-shift keying (OQPSK) modulation. The modulation scheme generates improper complex multiple-access interference (MAI) and necessitates the use of WL receivers. Focus is on bandlimited pulse shapes and the inherent cyclostationarity (CS) of the uncoded received signal. The enhanced WL receiver replaces the conventional chip-matched filter with new filters that exploit the CS of the received signal through frequency shifting. The proposed WL receiver is shown to outperform the classical strictly linear (SL) receiver when the interfering users are (quasi-)synchronous with respect to the user of interest. High-powered interfering users, that may exist to support high data rates, increase the performance improvement delivered by the WL receiver. Moreover, it is shown that MAI can become proper, either identically or asymptotically, when users are asynchronous and equally powered. This is despite the fact that individual interfering signals are improper. Numerical results demonstrate that the WL receiver can outperform the SL receiver by 1-3 dB under the examined scenarios with current CDMA standards settings. In asynchronous or quasi-synchronous transmission modes, performance gain of the WL receiver degrades unless the number of high-powered active users remains small. An example for implementation of the WL receiver is proposed and compared with that of the SL receiver when minimum-shift keying modulation, a special case of OQPSK, is used. The implementation is based on a fractionally spaced equalizer whose taps are updated by an adaptive algorithm. It is shown that the proposed structure is capable of delivering the maximum signal-to-noise ratio predicted by theory.

A new scheme for coding both the shape and texture of arbitrarily shaped visual objects is presented. Based on set partitioning on hierarchical trees (SPIHT), the proposed Shape and Texture SPIHT (ST-SPIHT) employs a novel implementation of the shape-adaptive discrete wavelet transform (SA-DWT) using in-place lifting, along with parallel coding of texture coefficients and shape mask pixels to create a single embedded code that allows for fine-grained rate-distortion scalability. The single output code simplifies the logistics of object storage and transmission compared to previously published schemes. An input parameter provides control over the relative progression between shape and texture coding in the embedded code, allowing for adjustment of the emphasis of shape versus texture quality in low bit rate reconstructions. The combination of features provided by ST-SPIHT, namely, explicit and progressive shape coding in parallel with wavelet-based embedded coding of the object texture, is unique compared to previously published schemes. Computational complexity is minimized since the shape coding takes advantage of the decomposition and spatial orientation trees used for texture coding. Objective and subjective simulation results show that the proposed ST-SPIHT scheme has rate-distortion performance comparable or superior to MPEG-4 Visual Texture Coding for most bit rates

In this paper, we propose a novel ensemble-based approach to boost performance of traditional Linear Discriminant Analysis (LDA)-based methods used in face recognition. The ensemble-based approach is based on the recently emerged technique known as "boosting". However, it is generally believed that boosting-like learning rules are not suited to a strong and stable learner such as LDA. To break the limitation, a novel weakness analysis theory is developed here. The theory attempts to boost a strong learner by increasing the diversity between the classifiers created by the learner, at the expense of decreasing their margins, so as to achieve a tradeoff suggested by recent boosting studies for a low generalization error. In addition, a novel distribution accounting for the pairwise class discriminant information is introduced for effective interaction between the booster and the LDA-based learner. The integration of all these methodologies proposed here leads to the novel ensemble-based discriminant learning approach, capable of taking advantage of both the boosting and LDA techniques. Promising experimental results obtained on various difficult face recognition scenarios demonstrate the effectiveness of the proposed approach. We believe that this work is especially beneficial in extending the boosting framework to accommodate general (strong/weak) learners.

The asymmetric signal constellation (ASC) method to break isometry is analyzed in a superimposed symbol framework with a Kalman filter estimator (KF) / maximum-likelihood (ML) detector as the receiver. Direct application of the ASC method led to a bit error floor, which motivates the proposal of combining orthogonal spreading codes with ASC to solve this problem. The proposed scheme generalizes previously proposed ASC and pilot-assisted solutions in a systematic way and results in coherent detection schemes without set bit error floors and better performance

A zooming framework suitable for single-sensor digital cameras is introduced and analyzed in this paper. The proposed framework is capable of zooming and enlarging data acquired by single-sensor cameras that employ the Bayer pattern as a color filter array (CFA). The approach allows for operations on noise-free data at the hardware level. Complexity and cost implementation are thus greatly reduced. The proposed zooming framework employs: 1) a spectral model to preserve spectral characteristics of the enlarged CFA image and 2) an adaptive edge-sensing mechanism capable of tracking the underlying structural content of the Bayer data. The framework readably unifies numerous solutions which differ in design characteristics, computational efficiency, and performance. Simulation studies indicate that the new zooming approach produces sharp, visually pleasing outputs and it yields excellent performance, in terms of both subjective and objective image quality measures.

The location of mobile terminals in cellular networks is an important problem with applications in resource allocation, location sensitive browsing, and emergency communications. Finding cost effective location estimation techniques that are robust to non-line of sight (NLOS) propagation, quantization, and measurement noise is a key problem in this area. Quantized time difference of arrival (TDoA) and received signal strength (RSS) measurements can be made simultaneously by CDMA cellular networks at low cost. The different sources of errors for each measurement type cause RSS and TDoA measurements to contain independent information about mobile terminal location. This paper applies data fusion to combine the information of RSS and TDoA measurements to calculate a superior location estimate. Nonparametric estimation methods, that are robust to variations of measurement noise and quantization, are employed to calculate the location estimates. It is shown how the data fusion location estimators are robust, provide lower error than the estimators based on the individual measurements, and have low implementation cost.

In many real applications for face recognition, such as surveillance photo identification, each subject only has one image sample for training which makes many supervised learning techniques fail to apply. Furthermore, since subject appearance has large variabilities due to aging, illumination and camera viewpoints, the face images to be identified are usually different from the stored templates. In this paper, a novel solution to this problem is proposed based on the well known unsupervised methodology, eigenface. We proposed a criterion to select the eigenfaces forming a feature subspace in which the intrapersonal variation is small compared to interpersonal variation and as well as most discriminating power is retained. The selection criterion maximizes the ratio between inter and intra personal variation, and at the same time takes total inter variation into account. Extensive experimentation following the FERET evaluation protocol indicates that the proposed scheme improves significantly the recognition performance.

The Bell-Labs layered space-time (BLAST) architecture is a simple and efficient multi-antenna coding structure that can achieve high spectral efficiency (Foschini, G. and Gans, M., Wireless Personal Commun., vol.6, p.311-35, 1998). Many BLAST detectors require more receiver antennas than transmitter antennas. We propose a novel turbo-processing BLAST detector based on a group detection strategy that can operate in systems with fewer receiver antennas than transmitter antennas. A maximum a posteriori (MAP) decision is made using a group of transmitted symbols and the remaining signal contribution is treated as interference. The interference is characterized as a non-zero mean colored noise source that is whitened before a decision is made. The proposed detector, the group MAP (GMAP) detector, is a generalization of both the MAP detector and the turbo-processing minimum mean squared error (MMSE) detector (Sellathurai, M. and Haykin, S., 2002; Abe, T. and Matsumoto, T., 2001). A novel grouping algorithm is proposed for the GMAP detector. Simulation is used to compare the GMAP detector with the MAP detector and MMSE detector.

This paper introduces a new color filter array (CFA) interpolation method for digital still cameras. The proposed interpolation scheme is able to overcome the limitations of existing CFA based image acquisition solutions, and restore color images without introducing false colors, edge blurring or visual artifacts. The method utilizes an edge-sensing mechanism, a color-difference model and a correction step. The employed correction process is useful in highly correlated images, however, in images with globally week correlation variations it is counter productive, therefore, a global control of the correction step is introduced. Simulation studies indicate that the proposed method is computationally efficient and yields excellent performance, in terms of subjective and objective image quality measures, while outperforming state-of-the-art CFA interpolation methods.

We propose a novel face detection algorithm which detects faces in color images using a combination of feature and decision fusion mechanisms. In addition to commonly used skin color information, two additional features, namely average face template matching score and horizontal edge template matching score are utilized. A mean shift algorithm operating on the combined feature space is used to determine face candidate areas. The face candidate and its flipped pattern are then inputted to a multiple layer perceptron based classifier. Two outputs along with the correlation value between the candidate and its flipped pattern are then combined to give the final decision. Experimentation on two different test databases indicates that the proposed method performs well under a variety of scale, expression and environmental conditions, outperforming commonly used approaches.

An open peer-to-peer architecture for performing distributed image indexing and retrieval is proposed. The system employs a sociological model of human acquaintance networks (small world theory) and concepts derived from the nature of the World Wide Web. Retrieval is performed using agents, and a node hopping algorithm is employed that exploits node referrals established from descriptor data stored locally by each node. A general framework for this small world image miner (SWIM) is presented along with a realization using MPEG-7 color structure descriptor data for 2400 images. Results related to search agent path length and network node degree are presented.

A new system is proposed for gait analysis and recognition applications. The new system is based on a denoising process and a new angular transform that are applied on binary silhouettes. Each human silhouette in a gait sequence is transformed into a low dimensional feature vector consisting of average pixel distances from the center of the silhouette. The sequence of feature vectors corresponding to a gait sequence is used for identification based on a minimum-distance criterion between test and reference sequences. By using the new system on the gait challenge database, improvements in recognition performance are seen in comparison to other methods of similar or higher complexity.

It is well-known that the applicability of linear discriminant analysis (LDA) to high-dimensional pattern classification tasks such as face recognition (FR) often suffers from the so-called "small sample size" (SSS) problem arising from the small number of available training samples compared to the dimensionality of the sample space. In this paper, we propose a new LDA method that effectively addresses the SSS problem using a regularization technique. In addition, a scheme of expanding the representational capacity of the face database is introduced to overcome the limitation that the LDA based algorithms require at least two samples per class available for learning. Extensive experimentation performed on the FERET database indicates that the proposed methodology outperforms traditional methods such as eigenfaces and direct LDA in a number of SSS setting scenarios.

A new bit-level based secret sharing scheme for encryption of private financial and pharmaceutical digital documents, and digital signature images is provided. The proposed {k,n} secret-sharing method allows for secret sharing of both scanned binary documents and the computer-generated artworks encrypting the document image into n shares. The secret information is recovered only if k (or more) allowed shares are available for decryption. Cryptographic operations for both encryption and decryption procedures are directly performed in the decomposed bit-level domain. The methods reveals the original document unchanged and thus, the scheme satisfies the perfect reconstruction property.

In this paper, a novel face recognition method is proposed for surveillance photo identification applications. In such a case, only a limited number of images per subject is available for training purposes. Furthermore, surveillance photos are usually different from the stored templates mostly due to aging, illumination and pose variations. It is common practice to apply unsupervised techniques such as principle component analysis (PCA) when the sample size for each subject is small. However, since PCA is performed without sample label considerations, the captured variation between images contains not only interpersonal variation but also intrapersonal variation which has an adverse effect on recognition performance. To overcome the problem, feature selection is performed in the PCA space to obtain a representation in which intrapersonal variation is minimized and interpersonal variation is maximized. Extensive experimentation following the FERET evaluation protocol indicates that the proposed scheme improves significantly the recognition performance.

We propose a methodology for gait recognition based on dynamic time warping. The gait sequences are initially partitioned into gait cycles and then the test cycles are compared to reference cycles using dynamic time warping. The final distance between a test and a reference sequence is determined using a nonlinear rule. Experimental results are reported showing an improvement in recognition performance in comparison to the baseline algorithm on the "gait challenge" database.

This work focuses on a camera image processing system suitable for cost-effective single-sensor imaging devices such as image-enabled wireless phones and pocket-size imaging devices. The color filter array (CFA) zooming, CFA interpolation and CFA based postprocessing steps, employed in the proposed system, are unified to use a simple linear interpolator defined over the Bayer CFA components. Using the spectral correlation characteristics expressed through the differences between the available color components, the proposed system eliminates color artifacts in the enlarged, full color camera output.

A new scheme suitable for simultaneous encryption and secret sharing of color images is presented. Employing a generalized {k, n}-secret sharing principle the method encrypts the color image input into n noise-like color shares. The secret information is recovered only if k allowed shares are available for decryption. To overcome limitations of traditional secret sharing schemes, the method i) operates at the bit-levels of the input and share images, ii) increases the level of protection utilizing full color gamut in generating the shares, and iii) satisfies the perfect reconstruction property, in terms of both the spatial resolution and the color/structural image content, for an authorized decryption process.

Location estimation of mobile users has been required in cellular networks for E-911 callers as a mandatory service. In this paper, radial basis function (RBF) neural networks with a hierarchical structure are proposed to estimate the mobile user location. At each level of hierarchical structure, the coverage area of base station is divided to different cells whose areas have overlaps. The RBF network of each cell is trained based on the received signals by an antenna array from the coverage area of that cell. The trained network estimates the mobile user location based on the power and arrival direction of the received signal. Mobile location estimation is improved by zooming in the smaller cell at each level of the hierarchical structure. Simulation results show that the proposed estimation method achieves a good performance with robustness in non-line of sight propagation and urban environments.

In multiplicative fading channels, joint channel estimation and data detection (CE/DD) schemes cannot differentiate among certain sequences of amplitude- and/or phase-modulated (AM/PM) symbols drawn from rotationally invariant signal constellations. This paper identifies these so-called isometric sequences as the main source of performance degradation, and introduces a unifying framework that effectively solves the problem by using asymmetric signal constellations (ASC) and a normalized innovations-based detector. The encompassing nature of the solution is clearly demonstrated by showing that seemingly unrelated previous results, such as training-based solutions, can be viewed as special cases of the modulation-based solution discussed here. A comprehensive analysis, supported by simulation studies, of the relationships among modulation schemes, isometry, and detection performance is provided. Results indicate that the proposed ASC solution offers excellent performance without incurring significant complexity or reducing the transmission rate. Furthermore, it is shown to be robust in various fading rates, and for different signal constellations.

From a perceptual standpoint, the subjectivity inherent in understanding and interpreting visual content in multimedia indexing and retrieval motivates the need for online interactive learning. Since efficiency and speed are important factors in interactive visual content retrieval, most of the current approaches impose restrictive assumptions on similarity calculation and learning algorithms. Specifically, content-based image retrieval techniques generally assume that perceptually similar images are situated close to each other within a connected region of a given space of visual features. This paper proposes a novel method for interactive image retrieval using query feedback. Query feedback learns the user query as well as the correspondence between high-level user concepts and their low-level machine representation by performing retrievals according to multiple queries supplied by the user during the course of a retrieval session. The results presented in this paper demonstrate that this algorithm provides accurate retrieval results with acceptable interaction speed compared to existing methods.

A postprocessing method for the correction of visual demosaicking artifacts is introduced. The restored, full-color images previously obtained by cost-effective color filter array interpolators are processed to improve their visual quality. Based on a localized color ratio model and the original underlying Bayer pattern structure, the proposed solution impressively removes false colors while maintaining image sharpness. At the same time, it yields excellent improvements in terms of objective image quality measures.

The large volumes of artistic visual data available to museums, art galleries, and online collections motivate the need for effective means to retrieve relevant information from such repositories. This paper proposes a decision making framework for content-based retrieval of art images based on a combination of low-level features. Traditionally, the similarity among two images has been calculated as a weighted distance between two feature vectors. This approach, however, may not be mathematically and computationally appropriate and does not provide enough flexibility in modeling user queries. This paper proposes a framework that generalizes a wide set of previous approaches to similarity calculation including the weighted distance approach. In this framework, image similarities are obtained through a decision making process based on low-level feature distances using fuzzy theory. The analysis and results of this paper indicate that the aggregation technique presented here provides an effective, general, and flexible tool for similarity calculation based on the combination of individual descriptors and features.

Linear Discriminant Analysis (LDA) is derived from the optimal Bayes classifier when classes are assumed to be Gaussian with identical covariance matrices. However, it is well known that the distribution of face images under a perceivable variation in viewpoint, illumination or facial ex-pression, is highly nonlinear and complex. The Quadratic Discriminant Analysis (QDA) which relaxes the identical covariance assumption and allows for nonlinear discrimi-nant boundaries to be formed, seems to be a better choice. However, the applicability of QDA to problems, such as face recognition, where the number of training samples is much smaller than the dimensionality of the sample space is problematic due to the increased number of parameters to be learned. In this paper, we propose a new regularized discriminant analysis method that effectively solves the so-called “small sample size ” problem in very high-dimensional face image space. Extensive experimentation performed on the FERET database indicates that the proposed method-ology outperforms traditional methods such as Eigenfaces, QDA and Direct LDA in a number of application scenarios. 1.

Adaptive nonlinear filtering methods are preferred in situations when it is necessary to adapt a filter behavior for varying signal and noise statistics. In the case of impulsive noise corruption, the problem is stated often as searching for the switching function that allows the filtering effect only to noisy samples. Thus, the undesired smoothing of non-corrupted image areas, which results in blurring, especially of small image structures and details, is reduced. We provide a new adaptive framework between a basic vector directional filter and an identity operation based on the angular multichannel definition of J.S. Lee's sigma filter (see Comp Vision, Graphics, Image Proc., vol.24, no.2, p.255-69, 1983).

The paper presents a new filtering scheme for the removal of impulsive noise in color images. It is based on estimating the probability density function for color pixels in a filter window by means of the kernel density estimation method. A quantitative comparison of the proposed filter with the vector median filter shows its excellent ability to reduce noise while simultaneously preserving fine image details.

In this paper, we propose a new algorithm to boost performance of traditional linear discriminant analysis (LDA)-based face recognition (FR) methods in complex FR tasks, where highly nonlinear face pattern distributions are often encountered. The algorithm embodies the principle of "divide and conquer", by which a complex problem, is decomposed into a set of simpler ones, each of which can be conquered by a relatively easy solution. The Ad-aBoost technique is utilized within this framework to: 1) generalize a set of simple FR sub-problems and their corresponding LDA solutions; 2) combine results from the multiple, relatively weak, LDA solutions to form a very strong solution. Experimentation performed on the FERET database indicates that the proposed methodology is able to greatly enhance performance of the traditional LDA-based method with an averaged improvement of correct recognition rate (CRR) up to 9% reported.

In this paper a novel approach to the problem of impulsive noise reduction in color images based on the nonparametric density estimation is presented. The basic idea behind the new image filtering technique is the maximization of the similarities between pixels in a predefined filtering window. The new method is faster than the standard vector median filter (VMF) and preserves better edges and fine image details. Simulation results show that the proposed method outperforms standard algorithms of the reduction of impulsive noise in color images.

This paper presents a new filtering scheme for the removal of impulsive noise in multichannel images. It is based on the estimation of the probability density function for pixels in a filtering window, by means of the kernel density estimation method. The filtering algorithm itself is based on the comparison of pixels with their neighborhood in a sliding window. The results obtained with the new filter show its excellent ability to reduce impulsive noise while simultaneously preserving fine image details.

In this paper we provide a new filtering scheme for the detection and the removal of impulsive noise in digital color images. The proposed adaptive nonlinear vector filters take the advantages of the robust order-statistic theory, generalized directional distance filter and standard sigma filter concept. The principles of the design are explained in detail. Simulation studies indicate that the proposed method is computationally attractive and is able to achieve excellent balance between the image-detail preservation and the noise attenuation.

In this paper a novel approach to the problem of impulsive noise removal in color images based on the nonparametric density estimation is presented. The basic idea behind the new image filtering technique is the maximization of the similarities between pixels in a predefined filtering window. The new method is faster than the standard vector median filter and preserves better edges and fine image details. Simulation results show that the proposed method outperforms other standard algorithms of the reduction of impulsive noise in color images.

Techniques that can introduce low-dimensional feature representation with enhanced discriminatory power is of paramount importance in face recognition (FR) systems. It is well known that the distribution of face images, under a perceivable variation in viewpoint, illumination or facial expression, is highly nonlinear and complex. It is, therefore, not surprising that linear techniques, such as those based on principle component analysis (PCA) or linear discriminant analysis (LDA), cannot provide reliable and robust solutions to those FR problems with complex face variations. In this paper, we propose a kernel machine-based discriminant analysis method, which deals with the nonlinearity of the face patterns' distribution. The proposed method also effectively solves the so-called "small sample size" (SSS) problem, which exists in most FR tasks. The new algorithm has been tested, in terms of classification error rate performance, on the multiview UMIST face database. Results indicate that the proposed methodology is able to achieve excellent performance with only a very small set of features being used, and its error rate is approximately 34% and 48% of those of two other commonly used kernel FR approaches, the kernel-PCA (KPCA) and the generalized discriminant analysis (GDA), respectively.

Low-dimensional feature representation with enhanced discriminatory power is of paramount importance to face recognition (FR) systems. Most of traditional linear discriminant analysis (LDA)-based methods suffer from the disadvantage that their optimality criteria are not directly related to the classification ability of the obtained feature representation. Moreover, their classification accuracy is affected by the "small sample size" (SSS) problem which is often encountered in FR tasks. In this paper, we propose a new algorithm that deals with both of the shortcomings in an efficient and cost effective manner. The proposed method is compared, in terms of classification accuracy, to other commonly used FR methods on two face databases. Results indicate that the performance of the proposed method is overall superior to those of traditional FR approaches, such as the eigenfaces, fisherfaces, and D-LDA methods.

Abstract Estimating the position of mobile terminals is an important problem for cellular networks. A low cost method of locating the mobile terminal is to use measurements of the radio path loss. The distribution of radio path loss is, unfortunately, a non‐linear function of the mobile terminal location. The non‐linearity results from large obstacles to radio‐wave propagation such as buildings or hills. This paper demonstrates how the conditional density of the location given measured path loss can be approximated as a sum of kernel density functions based on radio propagation data collected from propagation surveys or estimated from computer models. Using these approximate density functions an accurate location estimate of a mobile terminal can be estimated from measured path loss values contaminated by measurement noise. Copyright © 2002 John Wiley & Sons, Ltd.

The obstacle of generating hybrid queries within the context of content-based image retrieval is still very real. In attempts to overcome this, fuzzy aggregation can be used to combine single, simple index queries into larger, more complex ones. The paper outlines the use of a fuzzy aggregation technique for hybrid querying which has the ability to adjust its behavior according to operator-controlled parameters. The resulting aggregator can be viewed as a feature-adaptive overall similarity measure. We limit the scope of the aggregator to queries involving color content, color coverage, and horizontal/vertical trends, and apply it to a media database comprised of Corel images of fixed size. Preliminary results show promise and illustrate that hybrid queries using the fuzzy aggregator are effective in their ability to retrieve relevant images while suppressing erroneous retrievals when compared to simple, single-feature queries. In addition, the results obtained are at a minimum comparable to multiple-feature queries generated using a weighted mean approach, but exhibiting scalability and greater flexibility in parameter adjustment.

The performance of many state-of-the-art face recognition (FR) methods deteriorates rapidly when large databases are considered. We propose a novel clustering method based on a linear discriminant analysis methodology which deals with the problem of FR on a large-scale database. Contrary to traditional clustering methods such as K-means, which are based on certain "similarity criteria", the proposed method uses a novel "separability criterion" to partition a training set from the large database into a set of K smaller and simpler subsets or maximal-separability clusters (MSCs). Based on these MSCs, a novel two-stage hierarchical classification framework is proposed. Under the framework, the complex FR problem on a large database is decomposed into a set of simpler ones, where traditional methods can be successfully applied. Experiments with a database containing 1654 face images of 157 subjects indicate that the error rate performance of a traditional method under the proposed framework can be greatly improved without significantly increasing computational complexity.

Techniques that can introduce low-dimensional feature representation with enhanced discriminatory power is of paramount importance in face recognition applications. It is well known that the distribution of face images, under a perceivable variation in viewpoint, illumination or facial expression, is highly nonlinear and complex. It is therefore, not surprising that linear techniques, such as those based on principle component analysis (PCA) or linear discriminant analysis (LDA) cannot provide reliable and robust solutions to those complex face recognition problems. We propose a kernel machine based discriminant analysis method, which deals with the nonlinearity of the face patterns' distribution. The proposed method also effectively solves the "small sample size" (SSS) problem which exists in most face recognition tasks. The new algorithm has been tested, in terms of error rate performance, on the multi-view UMIST Face Database. Results indicate that the proposed methodology outperform other commonly used approaches, such as the kernel-PCA (KPCA) and the generalized discriminant analysis (GDA).

Recently there has been a lot of interest in improving the infrastructure used in medical applications. In particular, there is renewed interest on non-invasive, high-resolution diagnostic methods. One such method is digital, 3D ultrasound medical imaging. Current state-of-the-art ultrasound systems use specialized hardware for performing advanced processing of input data to improve the quality of the generated images. Such systems are limited in their capabilities by the underlying computing architecture and they tend to be expensive due to the specialized nature of the solutions they employ.Our goal in this work is twofold: (i) To understand the behavior of this class of emerging medical applications in order to provide an efficient parallel implementation and (ii) to introduce a new benchmark for parallel computer architectures from a novel and important class of applications. We address the limitations faced by modern ultrasound systems by investigating how all processing required by advanced beamforming algorithms can be performed on modern clusters of high-end PCs connected with low-latency, high-bandwidth system area networks. We investigate the computational characteristics of a state-of-the-art algorithm and demonstrate that today's commodity architectures are capable of providing almost-real-time performance without compromising image quality significantly.

We propose a new algorithm of noise reduction in color images. The new technique of multichannel image enhancement is capable of reducing impulse and Gaussian noise and it outperforms the basic methods based on vector median used for noise reduction in color images. A new smoothing operator, based on a random walk model and on a fuzzy similarity measure between pixels connected by a digital geodesic path is introduced. The efficiency of the proposed method was tested on the standard color images using the widely used objective image quality.

A color image segmentation technique is presented for use in coding and/or compression of video-conferencing sequences. The proposed technique utilizes the perceptual HSI (hue, saturation, intensity) color space and incorporates an unsupervised seed determination algorithm that finds the seed pixels that are in the spatial center of the regions in the image. The algorithm employs a hierarchy system where seeds found at the highest level produce the best results. The region growing algorithm uses these seeds to grow regions by appending to each seed pixel those neighboring pixels that satisfy a certain homogeneity criterion. The technique is found to be robust and relatively computationally inexpensive. The effectiveness of the algorithm is found to be much improved over techniques used in the past.

Processing multichannel signals using digital signal processing techniques has received increased attention lately due to its importance in applications such as multimedia technologies and telecommunications. The objective of this paper is twofold: 1) to introduce adaptive filtering techniques to the reader who is just beginning in this area and 2) to provide a review for the reader who may be well versed in signal processing. The perspective of the topic offered here is one that comes primarily from work done in the field of multichannel (color) image processing. Hence, many of the techniques and works cited here relate to image processing with the emphasis placed primarily on filtering algorithms based on fuzzy concepts, multidimensional scaling, and order statistics-based designs. It should be noted, however, that multichannel signal processing is a very broad field and thus contains many other approaches that have been developed from different perspectives, such as transform domain filtering, classical least-square approaches, neural networks, and stochastic methods, just to name a few. We present a general formulation based on fuzzy concepts, which allows the use of adaptive weights in the filtering structure, and we discuss different filter designs. The strong potential of fuzzy adaptive filters for multichannel signal applications, such as color image processing, is illustrated with several examples.

We present a technique for coarsely extracting the regions of natural color images which contain directional detail, e.g., edges, texture, etc., which we then use for image database indexing. As a measure of color activity, we use a perceptually modified distance measure based on the sum-of-angles criterion. We then apply histogram thresholding techniques to separate the image into smooth color regions and busy regions where edge, texture and colour activity exists. Database indices are then created from the busy regions using the directional detail histogram technique and retrieval is performed using these.

We address the issue of image database retrieval based on color using various vector distance metrics. Our system is based on color segmentation where only a few representative color vectors are extracted from each image and used as image indices. These vectors are then used with vector distance measures to determine similarity between a query color and a database image. We test numerous popular vector distance measures in our system and find that directional measures provide the most accurate and perceptually relevant retrievals.

The problem of narrowband interference suppression in impulsive noise environments is considered. To suppress the interference we introduce a new nonlinear estimator. The proposed estimator can be applied to any discrete time, linear system which is observed in additive non-Gaussian measurement noise. The new filter is recursive, computationally efficient and with significantly improved performance over other linear and nonlinear schemes which are currently used for interference suppression.

Color is the characteristic which is most used for image indexing and retrieval. Due to its simplicity, the color histogram remains the most commonly used method for color indexing and retrieval. However, the lack of good perceptual histogram similarity measures, the global color content of histograms and the erroneous retrieval results due to gamma nonlinearity, calls for improved methods. We implement a vector angular-based distance measure for image retrieval based on color. We build distance vectors in a multidimensional query space in which the retrieval ranking of each image is determined. Our system exhibits high flexibility by allowing all types of queries, including query by color, query by multiple colors and query by example. In addition, colors can be excluded in a query, without requiring an additional level of analysis.

Deconvolution is one of the most important aspects of seismic signal processing. The objective of the deconvolution procedure is to remove the obscuring effect of the wavelet's replica making up the seismic trace and therefore obtain an estimate of the reflection coefficient sequence. This paper introduces a new deconvolution algorithm. Optimal distributed estimators and smoothers are utilized in the proposed solution. The new distributed methodology, perfectly suitable for a multisensor environment, such as the seismic signal processing, is compared to the centralized approach, with respect to computational complexity and architectural efficiency. It is shown that the distributed approach greatly outperforms the currently used centralized methodology offering flexibility in the design of the data fusion network.

A new approach to the problem of time series classification is discussed. A new adaptive classification scheme is introduced and compared with existing approaches, such as the Bayesian approach and the incremental credit assignment approach. Simulation results are included to demonstrate the effectiveness of the new methodology.

New adaptive filters for color image processing are introduced and analyzed. The proposed adaptive methodology constitutes a unifying and powerful framework for multichannel signal processing. Using the proposed methodology, color image filtering problems are treated from a global viewpoint that readily yields and unifies previous, seemingly unrelated, results. The new filters utilize Bayesian techniques and nonparametric methodologies to adapt to local data in the color image. The principles behind the new filters are explained in detail. Simulation studies indicate that the new filters are computationally attractive and have excellent performance.

This paper addresses the problem of noise attenuation for multichannel data, such as color images. The proposed filter utilizes adaptive data dependent non-parametric techniques. Simulation results indicate that the new filter suppresses impulsive as well as Gaussian noise and preserves edges and details.

A non-linear digital filter is presented in this paper: this filter aims at extending the concept of scalar rank-ordering in the case of multichannel images. The filter is based on two steps: (1) a transformation from a p-dimensional steps to a one-dimensional space by means of a space filling curve; (2) a scalar median filtering step. Results which demonstrate the advantages and the good restoration computational performances of the filter are shown.

This paper addresses the problem of noise attenuation for multichannel data. The proposed filter utilizes adaptively determined data-dependent coefficients based on a novel distance measure which combines vector directional with vector magnitude filtering. The special case of color image processing is studied as an important example of multichannel signal processing.

The problem of estimating the return power in a LIDAR system in the presence of multiplicative noise (speckle) is addressed. A significant class of the partitioning approach is applied and comparisons are made with the extended Kalman filter (EKF) in the case where model parameter uncertainty exists. Through extensive simulations, the partitioned filter is shown to be significantly superior to the EKF algorithm.