Ensuring consistent quality in friction stir welding (FSW) remains challenging because defect formation and mechanical performance arise from complex thermo-mechanical interactions. This study presents an interpretable convolutional neural network-bidirectional long short-term memory (CNN–BiLSTM) framework for patch-level surface-defect detection and weld-level tensile-strength classification in ultrasonic vibration-assisted FSW AA2060-T8E30 joints. Surface images from 54 welds were divided into six ordered patches to preserve contextual information along the welding direction. DenseNet121 and VGG16 were evaluated as standalone CNNs and hybrid CNN–BiLSTM models. DenseNet121–BiLSTM achieved the best defect-detection performance, with 98.15% fixed-test accuracy and 95.54% mean leave-one-weld-out accuracy, compared with 92.59% and 90.23% for VGG16–BiLSTM. The DenseNet121–BiLSTM model was then applied to three-class tensile-strength classification relative to base-material ultimate tensile strength, achieving 88.89% fixed-test accuracy and 77.78% ± 7.41% under repeated stratified grouped cross-validation. Gradient-weighted Class Activation Mapping (Grad-CAM) visualisations identified surface regions associated with the predictions, while a graphical user interface integrated image input, classification, confidence reporting, and interpretation. The results demonstrate the feasibility of using standard weld-surface images for both visible defect detection and preliminary tensile-strength categorisation to support post-weld decision-making.

In this paper, a response biomarkers discovery framework based on discriminant analysis and causal inference is introduced. The framework has two main stages, causal bio-mining and bio-markers validation. At the causal bio-mining stage, the significant biomarkers are extracted from the randomized controlled trial (RCT) dataset by different techniques, discriminant analysis, feature ranking, statistical significance and association scoring. The extracted biomarkers are then assessed with respect to the treatment group classification, using causal inference propensity score matching. The causal biomarkers when applied to the subgroups classification provided better accuracy results, however using the minimum possible features, when their causal estimate is higher than 0.15 for both the treated and the control groups. The proposed framework’s efficacy was confirmed on two publicly available datasets: LiTMUS (GEO: GSE45484) and Breast Cancer (GEO: GSE20271). The performance of the framework was compared to established techniques, including those based on statistical variance and diagonal linear discriminant analysis (DLDA). The proposed framework demonstrably outperformed these benchmark methods. Using 3 features the Lithium subgroup classification accuracy is 83.33 %, while the Non-Lithium subgroup classification accuracy is 93.75 %, based on causal score>=0.2. Meanwhile, using 12 features the FAC×6 subgroup classification accuracy is 81.90 %, and using 13 features the T/FAC subgroup classification accuracy is 92.70 %, based on causal score >=0.15.

Understanding the drivers of well-being from longitudinal behavioural data is a fundamental challenge in biomedical informatics, where traditional analyses often conflate correlation with causation. This paper presents a rigorous application of causal inference to disentangle the drivers of well-being from complex longitudinal self-report data (N=141 enrolled; N=94 analysed after a priori completeness threshold of ≥20 of 28 daily entries). We introduce a novel computational metric, the Behaviour Self-Regulation Score (BSRS), to quantify both trait-like (long-term) and state-like (short-term) behavioural consistency from daily reports of physical activity and sleep. Employing causal graphical models and propensity score methods, we estimate the causal effects of these behavioural patterns, controlling for motivational and perceptual confounders. Our analysis uncovers distinct causal pathways: while long-term self-regulation (BSRS-L) has a stable positive causal effect, short-term behavioural consistency (BSRS-S) demonstrates a significantly stronger causal impact on daily well-being, despite a near-zero correlation. Furthermore, we demonstrate that features selected via our causal framework significantly improve the predictive accuracy of well-being in machine learning models compared to conventional feature selection methods. This work contributes a robust methodological framework for causal analysis of longitudinal self-report data and provides evidence that causally-informed modelling can identify more potent targets for digital health interventions.

Current non-destructive tomography sensing systems, particularly in the domain of eddy current sensing, lack the capability to dynamically and intelligently optimize sensing parameters in response to time-varying environments. To address this limitation, we propose a hybrid tomography sensing system that integrates physical artificial intelligence (PAI) with an electromagnetic dynamics model for intelligent sensing. This system combines a physical recurrent neural network (PRNN) entities performed by the programmable planar coil arrays with its digital counterpart in a closed-loop configuration, facilitating forward inference and the backpropagation of errors respectively. Additionally, the system leverages physics-guided methods based on electromagnetic field dynamics and employs a combination of standard neural network training techniques to optimize the parameters of PRNN, enabling real-time adaptive optimization of sensing. Theoretical modelling of the PRNN has been rigorously conducted in this study. Furthermore, high-fidelity electromagnetic tomography (EMT) results for non-destructive testing are demonstrated, showcasing the potential of physics-guided analogue AI in EMT sensing. The results shows that the electromagnetic controlling system, optimized through physical AI approach, achieves higher precision results with lower complexity compared to standard digital implementations in eddy current testing. This also provides a novel potential way for optimizing PNNs, thereby enhancing the acceleration of physical AI and projecting diversity of physical information into analogue AI solver.

Graph representation learning is fundamental for analyzing graph-structured data. Exploring invariant graph representations remains a challenge for most existing graph representation learning methods. In this paper, we propose a cross-view graph consistency learning (CGCL) method that learns invariant graph representations for link prediction. First, two complementary augmented views are derived from an incomplete graph structure through a coupled graph structure augmentation scheme. This augmentation scheme mitigates the potential information loss that is commonly associated with various data augmentation techniques involving raw graph data, such as edge perturbation, node removal, and attribute masking. Second, we propose a CGCL model that can learn invariant graph representations. A cross-view training scheme is proposed to train the proposed CGCL model. This scheme attempts to maximize the consistency information between one augmented view and the graph structure reconstructed from the other augmented view. Furthermore, we offer a comprehensive theoretical CGCL analysis. This paper empirically and experimentally demonstrates the effectiveness of the proposed CGCL method, achieving competitive results on graph datasets in comparisons with several state-of-the-art algorithms.

Apparent first impression prediction has made great progress with deep neural networks. There is a trend for multimodal fusion where features from different sources are fused together to improve the accuracy of the prediction. However, in a real-life scenario, it is often hard to gather features from different sources such as audio and background information. It is desirable to develop a method that could improve the prediction accuracy from a single source rather than multiple sources. This study developed a method to predict personality traits from a single source of information, i.e., facial information. Specifically, a pre-trained Deep Convolutional Neural Network was employed to extract emotional expression frame by frame in the video clip, which was then fed into a Long Short Term Memory model to predict the “Big Five” personality traits score. In Parallel, the model based on the static apparent facial features was trained, and finally, the facial feature and facial expression were fused with demographic data (age and gender). The proposed system is tested on the CharLearn Dataset and achieved an accuracy score of 90.67% ranked just below the top 5 CharLearn Competition. The result also showed that the dynamic emotional pattern has a positive impact on first impression prediction, especially on extraversion.

Neuroimaging techniques have had a major impact on medical science, allowing advances in the research of many neurological diseases and improving their diagnosis. In this context, multimodal neuroimaging approaches, based on the neurovascular coupling phenomenon, exploit their individual strengths to provide complementary information on the neural activity of the brain cortex. This work proposes a novel method for combining electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) to explore the functional activity of the brain processes related to low-level language processing of skilled and dyslexic seven-year-old readers. We have transformed EEG signals into image sequences considering the interaction between different frequency bands by means of cross-frequency coupling (CFC), and applied an activation mask sequence obtained from the local functional brain activity inferred from simultaneously recorded fNIRS signals. Thus, the resulting image sequences preserve spatial and temporal information of the communication and interaction between different neural processes and provide discriminative information that allows differentiation between controls and dyslexic subjects with an AUC of 77.1%. Finally, explainability is improved by introducing an easily comprehensible representation of the SHAP values obtained for the classification method in the brainSHAP maps.

In the realm of electromagnetic nondestructive testing (NDT), accurately identifying and characterizing flaws within various materials is crucial for ensuring structural integrity. This article proposes a novel intelligent electromagnetic perception framework that combines physical and digital artificial intelligence to address the sensitivity and accuracy limitations inherent in conventional electromagnetic NDT. Unlike traditional passive data acquisition methods, the proposed system integrates a physical electromagnetic neural network and a physics-aware reinforcement learning algorithm to adaptively optimize electromagnetic field sensing parameters in real-time, significantly enhancing sensitivity in regions close to defects. On the digital side, a sensor-informed diffusion model reconstructs high-resolution images from low-resolution optimal sensitivity sensor data, allowing for detailed analysis of defect contours and depths. Experimental results demonstrate a maximum sensitivity improvement of 105.8% and a minimum defect quantification of 0.2 mm, exceeding the performance of established electromagnetic NDT techniques. This innovative framework combines adaptive electromagnetic field focusing with advanced image reconstruction, establishing a new benchmark in real-time, high-precision defect detection. In addition, it is offering valuable applications in pipeline inspection, aerospace, and automotive industries.

Visual surface defect detection plays a crucial role in industrial quality control. A series of deep learning based algorithms have been introduced into visual defect detection, achieving remarkable performance. However, these algorithms generally suffer from inadequate adaptability to the expanding of detection categories, which limits their detection capabilities when dealing with complex and dynamic real-world application scenarios. To address this issue, this paper proposes an innovative incremental defect detection model. This model is based on learnable feature fusion and dynamic category modeling, aiming to enhance the model’s online learning and adaptability to new defect categories. To effectively utilize the existing data and optimize the model training process, this study introduces a macrodata probability control strategy to guide the reasonable configuration of training data as well as the model. Furthermore, to improve the performance during the detection process, an interactive feedback mechanism is constructed. This mechanism provides effective data in real-time during the detection process, driving the model to undergo dynamic training and gradually enhancing its recognition and adaptability to newly emerging defect types. To verify the practicality and effectiveness of the proposed algorithm, comprehensive comparative experiments were conducted on multiple datasets and a real-world detection platform was established for validation. The experimental results demonstrate the superior performance and dynamic adaptability in practical applications.

Long-term non-invasive sleep stage monitoring is instrumental in comprehending the progression of sleep disorders, cardiovascular diseases, and the interplay between sleep, type 2 diabetes, and neurodegenerative diseases. However, the conventional deep learning approach is susceptible to personal attributes (PAs) such as age, Body Mass Index, and severity of sleep apnea existing in the training dataset, potentially hindering its generalisation capacity to unseen cohorts. This paper introduces DSleepNet, a novel approach that disentangles the feature space into PA-specific and PA-agnostic components using two probabilistic encoders. The PA-agnostic features, designed to remain unaffected by personal attributes, outperformed the baseline CNN, improving the mean F1 score by up to 8.7% (baseline: 60.3) and Cohen's Kappa by 4.7% (baseline: 55.5), especially in reducing the impact of sleep apnea. DSleepNet functions without the need for target cohort data during training. It operates without the need to acquire PA data during inference, nor does it require fine-tuning. A novel Independent Excitation mechanism is incorporated into the latent feature space to remove correlations between the two types of features. Comprehensive testing in various PA settings has demonstrated its efficacy in improving the model's robustness.

Incomplete multiview clustering (IMVC) optimally integrates complementary information within incomplete multiview data to improve clustering performance. Several one-step graph-based methods show great potential for IMVC. However, the low-rank structures of similarity graphs are neglected at the initialization stage of similarity graph construction. Moreover, further investigation into complementary information integration across incomplete multiple views is needed, particularly when considering the low-rank structures implied in high-dimensional multiview data. In this paper, we present one-step adaptive graph learning (OAGL) that adaptively performs spectral embedding fusion to achieve clustering assignments at the clustering indicator level. We first initiate affinity matrices corresponding to incomplete multiple views using spare representation under two constraints, i.e., the sparsity constraint on each affinity matrix corresponding to an incomplete view and the degree matrix of the affinity matrix approximating an identity matrix. This approach promotes exploring complementary information across incomplete multiple views. Subsequently, we perform an alignment of the spectral block-diagonal matrices among incomplete multiple views using low-rank tensor learning theory. This facilitates consistency information exploration across incomplete multiple views. Furthermore, we present an effective alternating iterative algorithm to solve the resulting optimization problem. Extensive experiments on benchmark datasets demonstrate that the proposed OAGL method outperforms several state-of-the-art approaches.

Abstract During steel production, ensuring the integrity of the product’s surface is crucial for maintaining a competitive edge. Detecting surface flaws is a key component in preserving high standards of production quality, affecting both the end product’s reliability and the efficiency of the manufacturing process. Traditional methods for identifying these defects have relied heavily on manual inspection or basic computer vision techniques, both of which present a significant challenge in terms of accuracy and the health and safety of the inspectors. This paper explores an innovative approach by integrating a sequence generation model incorporating the transformer architecture, enhancing the detection of defects in the production of quality hot-rolled steel sheets. This method, which solves object detection through the lens of sequence generation, offers a more intricate analysis of images, enabling a comprehensive examination of surface anomalies along with detailed annotations regarding their nature and precise location. The approach not only proposes to increase the accuracy of defect detection but also highlights its adaptability for broader industrial uses.

We compare the outputs of several widely-utilized image quality metrics (such as the PSNR, SSIM, MS-SSIM, FSIMc & IMMSE) in the context of AI-mediated image super-resolution (SR) microscopy derived from a GAN. Although these metrics are often employed in the image analytics space for assaying image quality, the findings of our study indicate that such metrics may not be quite suitable for determining the quality of a GAN-generated image in the context of image SR microscopy, namely in the preservation of fine details and structures which are crucial aspects of high-resolution images (as determined visually). For instance, in some of the assayed images, we observed that these metrics returned a relatively favorable score, while on closer visual inspection, the generated image was observed to be prone to reconstruction artifacts, bearing little similarity to the Expected (ground truth) image. In this respect, we have sought to develop a custom image quality metric capable of assessing image similarity on a pixel-wise scale irrespective of the bit-depth of the image. Our proposed metric [termed the image comparative index (ICI)] has proven to be a viable determinant of similarity between 2 images, closely corroborating with a pixel-wise map of the differences between the assayed images, thereby allowing for more detailed analysis and identification of specific areas of improvement. We postulate that these results represent an important consideration for researchers seeking to utilize deep convolutional neural networks (DCNNs) for image SR (particularly when assessing the similarity between a DCNN-generated and the ground truth images during model training) with potential extrapolations of the proposed ICI metric in image classification & object detection use-cases as well.

Diabetes Mellitus (DM) represents a metabolic disorder characterized by consistently elevated blood glucose levels due to inadequate pancreatic insulin production. Type 1 DM (DM1) constitutes the insulin-dependent manifestation from disease onset. Effective DM1 management necessitates daily blood glucose monitoring, pattern recognition, and cognitive prediction of future glycemic levels to ascertain the requisite exogenous insulin dosage. Nevertheless, this methodology may prove imprecise and perilous. The advent of groundbreaking developments in information and communication technologies (ICT), encompassing Big Data, the Internet of Medical Things (IoMT), Cloud Computing, and Machine Learning algorithms (ML), has facilitated continuous DM1 management monitoring. This investigation concentrates on IoMT-based methodologies for the unbroken observation of DM1 management, thereby enabling comprehensive characterization of diabetic individuals. Integrating machine learning techniques with wearable technology may yield dependable models for forecasting short-term blood glucose concentrations. The objective of this research is to devise precise person-specific short-term prediction models, utilizing an array of features. To accomplish this, inventive modeling strategies were employed on an extensive dataset comprising glycaemia-related biological attributes gathered from a large-scale passive monitoring initiative involving 40 DM1 patients. The models produced via the Random Forest approach can predict glucose levels within a 30-minute horizon with an average error of 18.60 mg/dL for six-hour data, and 26.21 mg/dL for a 45-minute prediction horizon. These findings have also been corroborated with data from 10 Type 2 DM patients as a proof of concept, thereby demonstrating the potential of IoMT-based methodologies for continuous DM monitoring and management. The integration of innovative biological signal sensors and the application of transformative trends in ICT can offer a novel perspective on DM treatment, ensuring precise and secure glucose level management.

Concept drift is a prevalent phenomenon in data streams that necessitates detection and in-depth understanding, as it signifies that the statistical properties of a target variable, which the model aims to predict, change over time in unforeseen ways. Existing detection methods predominantly aim to identify the drift start time, which lack comprehensive understanding of data streams, leading to a loss of drift information. In this paper, we present a novel Quadruple-based Approach for Understanding Concept Drift in Data Streams (QuadCDD) framework that not only detects and predicts the concept drift start point but also offers a more detailed analysis of concept drift through the use of quadruples, encompassing drift start, drift end, drift severity, and drift type. Our framework employs quadruples to enable informed decision-making and adopt appropriate actions to handle various concept drifts, effectively maintaining high and stable performance in data streams with concept drift. Experimental results validate the effectiveness of our QuadCDD framework in accurately detecting and understanding concept drifts, as well as in preserving the stability and performance of models in the presence of these drifts.

The tensor nuclear norm (TNN), defined as the sum of nuclear norms of frontal slices of the tensor in a frequency domain, has been found useful in solving low-rank tensor recovery problems. Existing TNN-based methods use either fixed or data-independent transformations, which may not be the optimal choices for the given tensors. As the consequence, these methods cannot exploit the potential low-rank structure of tensor data adaptively. In this article, we propose a framework called self-adaptive learnable transform (SALT) to learn a transformation matrix from the given tensor. Specifically, SALT aims to learn a lossless transformation that induces a lower average-rank tensor, where the Schatten- p quasi-norm is used as the rank proxy. Then, because SALT is less sensitive to the orientation, we generalize SALT to other dimensions of tensor (SALTS), namely, learning three self-adaptive transformation matrices simultaneously from given tensor. SALTS is able to adaptively exploit the potential low-rank structures in all directions. We provide a unified optimization framework based on alternating direction multiplier method for SALTS model and theoretically prove the weak convergence property of the proposed algorithm. Experimental results in hyperspectral image (HSI), color video, magnetic resonance imaging (MRI), and COIL-20 datasets show that SALTS is much more accurate in tensor completion than existing methods. The demo code can be found at https://faculty.uestc.edu.cn/gaobin/zh_CN/lwcg/153392/list/index.htm.

This paper presents novel modifications to the Firefly Algorithm (FA) that manipulate the functionality of the intensity and attractiveness of fireflies through the incorporation of grouping behaviours into the movement of the fireflies. FA is one of the most well-known and actively researched swarm-based algorithms, gaining notoriety for the powerful search capability offered and overall computational simplicity. While the FA is an effective optimisation algorithm, it is unfortunately susceptible to the issue of premature convergence and oscillations within the swarm, which can lead to suboptimal performance. In the original FA formulation, at each iteration fireflies will instinctively move towards the most intensely bright firefly which is in closest proximity to them. The algorithm proposed in this paper manipulates the movement of the fireflies through modification of this intensity and attraction relationship, allowing the swarm to move in different ways, ultimately increasing the search diversity within the swarm. While group-based FAs have been proposed previously, the group-based FAs presented in this paper utilise a different approach to creating groups, implementing groupings based upon firefly performance at each iteration, resulting in continually varying groupings of fireflies, to further increase search diversity and maintain computational simplicity.

Multiview clustering (MVC) aims to reveal the underlying structure of multiview data by categorizing data samples into clusters. Deep learning-based methods exhibit strong feature learning capabilities on large-scale datasets. For most existing deep MVC methods, exploring the invariant representations of multiple views is still an intractable problem. In this paper, we propose a cross-view contrastive learning (CVCL) method that learns view-invariant representations and produces clustering results by contrasting the cluster assignments among multiple views. Specifically, we first employ deep autoencoders to extract view-dependent features in the pretraining stage. Then, a cluster-level CVCL strategy is presented to explore consistent semantic label information among the multiple views in the fine-tuning stage. Thus, the proposed CVCL method is able to produce more discriminative cluster assignments by virtue of this learning strategy. Moreover, we provide a theoretical analysis of soft cluster assignment alignment. The extensive experimental results obtained on several datasets demonstrate that the proposed CVCL method outperforms several state-of-the-art approaches.

Convolutional neural networks (CNN) have been utilized to detect anomalies on the railway track surfaces whose conditions must be monitored to ensure the safety of railway systems. While CNN has advantages over conventional image processing methods in self-learning features for detecting railway track anomalies, the CNN model and parameters still need to be carefully constructed and examined for the effective application with railway track images. This study presents a systematic investigation of CNN model parameters for detecting anomalies on railway tracks. Parameters such as number of convolutional layers, convolutional kernel size, pooling kernel size and number of epochs were examined. Experiments and analyses were performed to determine how these parameters affect the detection accuracy. The experimental procedures and findings demonstrated the effects of individual parameters, as well as the potential interactions between the factors; thus, systematic procedures are needed to investigate and improve CNN models deployed to detect and classify anomalies on railway tracks.

Abstract Human activity recognition based on generated sensor data plays a major role in a large number of applications such as healthcare monitoring and surveillance system. Yet, accurately recognizing human activities is still challenging and active research due to people’s tendency to perform daily activities in a different and multitasking way. Existing approaches based on the recurrent setting for human activity recognition have some issues, such as the inability to process data parallelly, the requirement for more memory and high computational cost albeit they achieved reasonable results. Convolutional Neural Network processes data parallelly, but, it breaks the ordering of input data, which is significant to build an effective model for human activity recognition. To overcome these challenges, this study proposes causal convolution based on performers-attention and supervised contrastive learning to entirely forego recurrent architectures, efficiently maintain the ordering of human daily activities and focus more on important timesteps of the sensors’ data. Supervised contrastive learning is integrated to learn a discriminative representation of human activities and enhance predictive performance. The proposed network is extensively evaluated for human activities using multiple datasets including wearable sensor data and smart home environments data. The experiments on three wearable sensor datasets and five smart home public datasets of human activities reveal that our proposed network achieves better results and reduces the training time compared with the existing state-of-the-art methods and basic temporal models.

While the brain connectivity network can inform the understanding and diagnosis of developmental dyslexia, its cause-effect relationships have not yet enough been examined. Employing electroencephalography signals and band-limited white noise stimulus at 4.8 Hz (prosodic-syllabic frequency), we measure the phase Granger causalities among channels to identify differences between dyslexic learners and controls, thereby proposing a method to calculate directional connectivity. As causal relationships run in both directions, we explore three scenarios, namely channels' activity as sources, as sinks, and in total. Our proposed method can be used for both classification and exploratory analysis. In all scenarios, we find confirmation of the established right-lateralized Theta sampling network anomaly, in line with the assumption of the temporal sampling framework of oscillatory differences in the Theta and Gamma bands. Further, we show that this anomaly primarily occurs in the causal relationships of channels acting as sinks, where it is significantly more pronounced than when only total activity is observed. In the sink scenario, our classifier obtains 0.84 and 0.88 accuracy and 0.87 and 0.93 AUC for the Theta and Gamma bands, respectively.

Concept drift refers to the inevitable phenomenon that influences the statistical features of the data stream. Detecting concept drift in data streams quickly and precisely remains challenging, and failure to detect it will render model trained on historical data ineffective. Current drift detection methods suffer from the following problems: detection delay, many false detected drifts, and rarely utilize the deep neural network directly in the field of concept drift detection, which is considerably competent at addressing the classification problems of data stream. Furthermore, the output of the model is usually taken as a metric to detect drift, while changes in the model parameters are often ignored which contain highly useful information. In this paper, we propose a model-centric framework for concept drift detection that uses deep neural network to detect concept drift by focusing on the change in the model itself, rather than the model output. In addition, transfer learning is developed to accelerate the drift detection process and reduce the computational complexity by freezing parts of the network. To further reduce false detected drifts, we propose long and short time windows method to determine the real drift from the potential detected drift. Experiments with real-world and artificial datasets have been undertaken to demonstrate the effectiveness of the proposed framework.

As pipelines take an increasingly important role in energy transportation, their health management is necessary. In-pipe inspection is a common pipeline life maintenance method. The signal obtained through internal inspection contains strong noise and interference where the internal environment of the pipeline is extremely complicated. Thus, it is challenging to accurately identify the defect signal. In this article, a defect detection framework based on feature boosting is proposed by using the multisensing pipeline pig as the detection signals. Through boosting construction of features and hierarchical classification, the framework can not only correctly classify various signals in the internal detection signals but also realize the accurate identification of defect signals. Concurrently, in order to demonstrate the high flexibility and robustness of the detection framework, experiments, and verifications have been carried out on specimens in three different environments, i.e., 1) laboratory environment, 2) simulated environment, and 3) actual environment. In the classification of actual environmental detection signals, quantitative evaluation with different algorithms have been undertaken using the F-score to demonstrate the effectiveness of the proposed framework.

Schizophrenia is a severe mental health condition that requires a long and complicated diagnostic process. However, early diagnosis is vital to control symptoms. Deep learning has recently become a popular way to analyse and interpret medical data. Past attempts to use deep learning for schizophrenia diagnosis from brain-imaging data have shown promise but suffer from a large training-application gap — it is difficult to apply lab research to the real world. We propose to reduce this training-application gap by focusing on readily accessible data. We collect a data set of psychiatric observations of patients based on DSM-5 criteria. Because similar data is already recorded in all mental health clinics that diagnose schizophrenia using DSM-5, our method could be easily integrated into current processes as a tool to assist clinicians, whilst abiding by formal diagnostic criteria. To facilitate real-world usage of our system, we show that it is interpretable and robust. Understanding how a machine learning tool reaches its diagnosis is essential to allow clinicians to trust that diagnosis. To interpret the framework, we fuse two complementary attention mechanisms, ‘squeeze and excitation’ and ‘self-attention’, to determine global attribute importance and attribute interactivity, respectively. The model uses these importance scores to make decisions. This allows clinicians to understand how a diagnosis was reached, improving trust in the model. Because machine learning models often struggle to generalise to data from different sources, we perform experiments with augmented test data to evaluate the model’s applicability to the real world. We find that our model is more robust to perturbations, and should therefore perform better in a clinical setting. It achieves 98% accuracy with 10-fold cross-validation.

This article presents a novel semisupervised learning method for wearable sensors to recognize human activities. The proposed method is termed a tri-very fast decision tree (VFDT). The proposed method is a more efficient version of the Hoeffding tree and three VFDTs are generated from the original labeled example set and refined using unlabeled examples. Based on the heuristic growth characteristics of VFDT, a tri-training framework is proposed which uses unlabeled data to update the model without labeled data. This significantly reduces the computational time and storage of the data processing. In addition, the proposed method is embedded into wearable devices for online learning, while the test data flow is regarded as the unlabeled data to update the model. The experiment collects data stream of 16 min with motion state switching frequently while the wearable devices recognize motions in real time. An experimental comparison has also been undertaken for performance evaluation between the wearable and computation using a desktop computer. The obtained results show that only minor difference in terms of the$f1$-score rendered by the proposed method online or offline. This is a prominent characteristic for wearable computing within the Internet of Things (IoT). Data set can be linked ashttps://faculty.uestc.edu.cn/gaobin/zh_CN/lwcg/153392/list/index.htm.

From the sparsity of vector to the sparsity of singular values, which essentially characterizes the low rank property of matrix. The sparsity measure based model is of significant interest in a range of contemporary applications in data analysis. However, there are different measurement strategies for sparse characterization of high dimensional tensor data. Albeit, most of the existing sparsity measures only consider the number of non-zero factor components, but ignore the geometric position distribution structure of non-zero elements in high-dimensional space. In this paper, based on the fact that sliced sparse distribution of the core tensor, a novel high order structure sparsity measure is proposed. More specifically, the sparsity measure unifies Tucker and CP tensor decomposition into a framework for general tensor. The CP decomposition of the core tensor with factor group sparse constraint realizes modeling the global low CP rank and the sliced sparse distribution of the non-zeros elements of the core tensor simultaneously. We apply minimizing high order structure sparse measurement to multispectral image denoising and deduce Alternating Direction Method of Multipliers (ADMM) optimization method to solve the model effectively. The subsequent experimental results show that the proposed algorithm is competitive with state-of-the art denoising methods.1

Abstract Systems of sensor human activity recognition are becoming increasingly popular in diverse fields such as healthcare and security. Yet, developing such systems poses inherent challenges due to the variations and complexity of human behaviors during the performance of physical activities. Recurrent neural networks, particularly long short-term memory have achieved promising results on numerous sequential learning problems, including sensor human activity recognition. However, parallelization is inhibited in recurrent networks due to sequential operation and computation that lead to slow training, occupying more memory and hard convergence. One-dimensional convolutional neural network processes input temporal sequential batches independently that lead to effectively executed operations in parallel. Despite that, a one-dimensional Convolutional Neural Network is not sensitive to the order of the time steps which is crucial for accurate and robust systems of sensor human activity recognition. To address this problem, we propose a network architecture based on dilated causal convolution and multi-head self-attention mechanisms that entirely dispense recurrent architectures to make efficient computation and maintain the ordering of the time steps. The proposed method is evaluated for human activities using smart home binary sensors data and wearable sensor data. Results of conducted extensive experiments on eight public and benchmark HAR data sets show that the proposed network outperforms the state-of-the-art models based on recurrent settings and temporal models.

This article proposes a fuzzy logic-based energy-management system (FEMS) for a grid-connected microgrid with renewable energy sources (RESs) and energy storage system (ESS). The objectives of the FEMS are reducing the average peak load (APL) and operating cost through arbitrage operation of the ESS. These objectives are achieved by controlling the charge and discharge rate of the ESS based on the state of charge of ESS, the power difference between load and RES, and electricity market price. The effectiveness of the fuzzy logic greatly depends on the membership functions (MFs). The fuzzy MFs of the FEMS are optimized offline using a Pareto-based multiobjective evolutionary algorithm, nondominated sorting genetic algorithm (NSGA-II). The best compromise solution is selected as the final solution and implemented in the fuzzy-logic controller. A comparison with other control strategies with similar objectives is carried out at a simulation level. The proposed FEMS is experimentally validated on a real microgrid in the energy storage test bed at Newcastle University, U.K.

With the increasing use of induction thermography (IT) for nondestructive testing in the mechanical and rail industry, it becomes necessary for the Manufacturers to rapidly and accurately monitor the health of specimens. The most general problem for IT detection is due to strong noise interference. In order to counter it, general postprocessing is carried out. However, due to the more complex nature of noise and irregular shape specimens, this task becomes difficult and challenging. In this article, a low-rank tensor with a sparse mixture of Gaussian (LRTSMoG) decomposition algorithm for natural crack detection is proposed. The proposed algorithm models jointly the LRST pattern by using a tensor decomposition framework. In particular, the weak natural crack information can be extracted from strong noise. Low-rank tensor based iterative sparse MoG noise modeling is carried out to enhance the weak natural crack information as well as reducing the computational cost. In order to show the robustness and efficacy of the model, experiments are conducted for natural crack detection on a variety of specimens. A comparative analysis is presented with general tensor decomposition algorithms. The algorithms are evaluated quantitatively based on signal-to-noise-ratio along with the visual comparative analysis.

Eddy current testing (ET) and eddy current thermography (ECT) are both important nondestructive testing methods that have been widely used in the field of conductive materials evaluation. Conventional ECT systems have often employed to test static specimens even though they are inefficient when the specimen is large. In addition, the requirement of high-power excitation sources tends to result in bulky detection systems. To mitigate these problems, in this article, a moving detection mode of multiphysics structured ET and ECT is proposed in which a novel L-shape ferrite magnetic yoke circumambulated with array coils is designed. The theoretical derivation model of the proposed method is developed which is shown to improve the detection efficiency without compromising the excitation current by ECT. The specimens can be speedily evaluated by scanning at a speed of 50-250 mm/s while reducing the power of the excitation current due to the supplement of ET. The unique design of the excitation-receiving structure has also enhanced the detectability of omnidirectional cracks. Moreover, it does not block the normal direction visual capture of the specimens. Both numerical simulations and experimental studies on different defects have been carried out and the obtained results have shown the reliability and detection efficiency of the proposed system.

This article proposes a hybrid multi-dimensional features fusion structure of spatial and temporal segmentation model for automated thermography defects detection. In addition, the newly designed attention block encourages local interaction among the neighboring pixels to recalibrate the feature maps adaptively. A Sequence-PCA layer is embedded in the network to provide enhanced semantic information. The final model results in a lightweight structure with smaller number of parameters and yet yields uncompromising performance after model compression. The proposed model allows better capture of the semantic information to improve the detection rate in an end-to-end procedure. Compared with current state-of-the-art deep semantic segmentation algorithms, the proposed model presents more accurate and robust results. In addition, the proposed attention module has led to improved performance on two classification tasks compared with other prevalent attention blocks. In order to verify the effectiveness and robustness of the proposed model, experimental studies have been carried out for defects detection on four different datasets. The demo code of the proposed method can be linked soon: http://faculty.uestc.edu.cn/gaobin/zh_CN/lwcg/153392/list/index.htm.

This paper presents the design of a fuzzy logic-based controller to be embedded in a grid-connected microgrid with renewable and energy storage capability. The objectives of the controller is to control the charge and discharge rate of the energy storage system (ESS) to reduce the end-user operating cost through arbitrage operation of the ESS and reducing the power exchange between the main and microgrid. Instead of using a forecasting-based approach, the proposed methodology takes the difference between the available renewable generation and load, state-of-charge of ESS, and electricity market price to determine the charge and discharge rate of the ESS in a rolling horizon. A comparison with other controllers with similar objectives shows that the proposed controller can achieve a lower operating cost and reduce the power exchange between the main and microgrid.

The remote monitoring of electrical systems has progressed beyond the need of knowing how much energy is consumed. As the maintenance procedure has evolved from reactive to preventive to predictive, there is a growing demand to know what appliances reside in the circuit (classification) and a need to know whether any appliance requires attention and maintenance (anomaly warning). Targeting at the increasing penetration of dc appliances and equipment in households and offices, the described low-cost solution consists of multiple distributed slave meters with a single master computer for extra low voltage dc picogrids. The slave meter acquires the current and voltage waveform from the cable of interest, conditions the data, and extracts four features per window block that are sent remotely to the master computer. The proposed solution uses a hierarchical extended k-nearest neighbors (HE-k NNs) technique that exploits the use of distance in k NN algorithm and considers a window block instead of individual data point for classification and anomaly warning to trigger the attention of the user. This solution can be used as an ad hoc standalone investigation of suspicious circuit or further expanded to several circuits in a building or vicinity to monitor the network. The solution can also be implemented as part of an Internet of Things application. This article presents the successful implementation of the HE-k NN technique in three different circuits: 1) lighting; 2) air-conditioning; and 3) multiple load dc picogrids with accuracy of over 93%. Its performance is superior over other anomaly warning techniques with the same set of data.

In the modern smart home, smart meters, and Internet of Things (IoT) have been massively deployed to replace traditional analogue meters. It digitalises the data collection and the meter readings. The data can be wirelessly transmitted that significantly reduces manual works. However, the community of smart home network is vulnerable to energy theft. Such attacks cannot be effectively detected since the existing techniques require certain devices to be installed to work. This imposes a challenge for energy theft detection systems to be implemented despite the lack of energy monitoring devices. This paper develops an energy detection system called smart energy theft system (SETS) based on machine learning and statistical models. There are three stages of decision-making modules, the first stage is the prediction model which uses multimodel forecasting system. This system integrates various machine learning models into a single forecast system for predicting the power consumption. The second stage is the primary decision making model that uses simple moving average (SMA) for filtering abnormally. The third stage is the secondary decision making model that makes the final stage of the decision on energy theft. The simulation results demonstrate that the proposed system can successfully detect 99.96% accuracy that enhances the security of the IoT-based smart home.

Long-term wellbeing monitoring is an underlying theme for evaluating health status by collecting physiological signs through behavioral traits. In alignment with Internet of Things (IoT), nonintrusive and trustworthy wearable social sensing technology holds a potential way for researchers to find and establish the interrelationships between unobtrusive social cues and physical mental health. This paper implements an IoT structured wearable social sensing platform with the integration of privacy audio feature, behavior monitoring, and environment sensing in a naturalistic environment. Particularly, four privacy protected audio-wellbeing features are embedded into the platform to automatically evaluate speech information without preserving raw audio data. Four weeks of long-term monitoring experimental studies have been conducted. A series of well-being questionnaires in conjunction with a group of students are engaged to objectively investigate the relationships between physical and mental health by utilizing the feature fusion strategy from speech, behavioral activities, and ambient factors.

A modified multiple generalized regression neural network (GRNN) is proposed to predict the noise level of various compartments onboard of the offshore platform. With limited samples available during the initial design stage, GRNN can cause errors when it maps the available inputs to sound pressure level for the entire offshore platform. To obtain more relevant group for GRNNs training, fuzzy C-mean (FCM) is used. However, outliers in some group may interfere the prediction accuracy. The problem of selecting suitable inputs parameters (in each cluster) is often impeded by lack of accurate information. Principal component analysis (PCA) is used to ensure high relevance input variables in each cluster. By fusing multiple GRNNs by an optimal spread parameter, the proposed modeling scheme becomes quite effective for modeling multiple frequency-dependent data set (ranging from 125 to 8000 Hz) with different input parameters. The performance of FCM-PCA-GRNNs has improved significantly as the results show a 25% improvement on the spatial sound pressure level (SPL) and 85% improvement on the spatial average SPL than just GRNNs alone. By comparing with data obtained from real engine room on a jack-up rig, the FCM-PCA-GRNNs noise model performs better with around 16% less error than the empirical-based acoustic models. Additionally, the results show comparable performance to statistical energy analysis that requires more time and resources to solve during the early stage of the offshore platform design.

With the increasing importance of using carbon fiber reinforced polymer (CFRP) composite in the aircraft industry, it becomes ever more critical to monitor the quality and health of CFRP during the manufacturing process as well as the in-service procedure. The most common types of defects in the CFRP are debonds and delaminations. It is difficult to detect the inner defects on a complex-shaped specimen using conventional nondestructive testing (NDT) methods. In this paper, an unsupervised machine learning method based on wavelet-integrated alternating sparse dictionary matrix decomposition is proposed to extract the weaker and deeper defect information for CFRP by using the optical pulse thermography (OPT) system. We propose to model the low-rank and sparse decomposition jointly in an alternating manner. By incorporating the low-rank information into the sparse matrix and vice versa, the weaker defects will be more efficiently extracted from noise and background. In addition, the integration of wavelet analysis with dictionary factorization enables an efficient time-frequency mining of information and significantly removes the high frequency noise as well as boosts the speed of computations. To investigate the efficacy and robustness of the proposed method, experimental studies have been carried out for inner debond defects on both regular- and irregular-shaped CFRP specimens. A comparative analysis has also been undertaken to study the proposed method against the general OPTNDT methods. The MATLAB demo code can be linked: http://faculty.uestc.edu.cn/gaobin/zh_CN/lwcg/153392/list/index.htm.

This paper presents an adaptive heartrate-dependent heartwave-signal-based biometric identification. A reliable and continuous heartwave extraction method featuring the hybridized discrete waveform transform method with heartrate adaptive QT and PR intervals to perform comprehensive heartwave features extractions on more than 35 000 heartwave signal. The size of training data was determined and the hybridized Gaussian-mixture-model-hidden-Markov-model classification method was used in the classification. Dynamic thresholding criterial incorporating user-specific scores and heartrate were adopted. The identification process using dynamic thresholding criterial achieved a remarkable receiver operating characteristic of 0.89 in true positive rate and an equal error rate of 0.11.

This paper presents a heartwave based authentication method that utilizes an ensemble of deep belief networks (DBNs) under different parameters to increase the reliability of feature extraction. The multiview outputs are further embedded into a single view before inputting into a stacked DBN for classification. The result of the proposed novel architecture achieved a classification rate of 98.3% with 30% training data. Importantly, it is able to perform user classification using heartwave signals acquired under intense physical exercise where heart rate ranges from 50 bpm to as high as 180 bpm. Under extreme physical duress, the heartwave from an individual experiences extreme morphological variations that render conventional classification approaches nonapplicable.

A novel unsupervised sparse component extraction algorithm for diagnosing micro defects in thermography imaging system is presented. The approach is optimized under Variational Bayesian framework, which is fully automated and does not require manual selection of the parameters in the solution. An internal sub sparse grouping mechanism and adaptive fine-tuning have been built into the proposed algorithm to control the sparsity. The proposed method is used to automatically detect the micro defects on metals. Other contending defect feature extraction and sparse pattern extraction methods are employed for comparison. The algorithm has been shown to improve the detection precision of both artificial and natural cracks.

Finger texture (FT) images acquired from different spectral lighting sensors reveal various features. This inspires the idea of establishing a recognition model between FT features collected using two different spectral lighting forms to provide high recognition performance. This can be implemented by establishing an efficient feature extraction and effective classifier, which can be applied to different FT patterns. So, an effective feature extraction method called the surrounded patterns code (SPC) is adopted. This method can collect the surrounded patterns around the main FT features. It is believed that these patterns are robust and valuable. Furthermore, a novel classifier termed the re‐enforced probabilistic neural network (RPNN) is proposed. It enhances the capability of the standard PNN and provides better recognition performance. Two types of FT images from the multi‐spectral Chinese Academy of Sciences Institute of Automation (CASIA) database were employed as two types of spectral sensors were used in the acquiring device: the white (WHT) light and spectral 460 nm of blue (BLU) light. Supporting comparisons were performed, analysed and discussed. The best results were recorded for the SPC by enhancing the equal error rates at 4% for spectral BLU and 2% for spectral WHT. These percentages have been reduced to 0% after utilising the RPNN.

This study proposes an optimal design of the orders of the discrete fractional Fourier transforms (DFrFTs) and construct an overcomplete transform using the DFrFTs with these orders for performing the sparse representations. The design problem is formulated as an optimisation problem with an ‐norm non‐convex objective function. To avoid all the orders of the DFrFTs to be the same, the exclusive OR of two constraints are imposed. The constrained optimisation problem is further reformulated to an optimal frequency sampling problem. A method based on solving the roots of a set of harmonic functions is employed for finding the optimal sampling frequencies. As the designed overcomplete transform can exploit the physical meanings of the signals in terms of representing the signals as the sums of the components in the time–frequency plane, the designed overcomplete transform can be applied to many applications.

Internet of Things makes deployment of smart home concept easy and real. Smart home concept ensures residents to control, monitor, and manage their energy consumption without any wastage. This paper presents a self-learning home management system. In the proposed system, a home energy management system, demand side management system, and supply side management system were developed and integrated for real time operation of a smart home. This integrated system has some capabilities such as price forecasting, price clustering, and power alert system to enhance its functions. These enhancing capabilities were developed and implemented using computational and machine learning technologies. In order to validate the proposed system, real-time power consumption data was collected from a Singapore smart home and a realistic experimental case study was carried out. The case study has shown that the developed system has performed well and created energy awareness to the residents. This proposed system also displays its ability to customize the model for different types of environments compared to traditional smart home models.

This paper presents a novel computational intelligence vision sensing approach to estimate nutrient content in wheat leaves by analyzing color features of the leaves images captured on field with various lighting conditions. We propose the development of deep sparse extreme learning machines (DSELM) fusion and genetic algorithm (GA) to normalize plant images as well as to reduce color variability due to a variation of sunlight intensities. We also apply the DSELM in image segmentation to differentiate wheat leaves from a complex background. In this paper, four moments of color distribution of the leaves images (mean, variance, skewness, and kurtosis) are extracted and utilized as predictors in the nutrient estimation. We combine a number of DSELMs with committee machine and optimize them using the GA to estimate nitrogen content in wheat leaves. The results have shown the superiority of the proposed method in the term of quality and processing speed in all steps, i.e., color normalization, image segmentation, and nutrient prediction, as compared with other existing methods.

Reciprocating impact load leads to plastic deformation on the surface of the kinematic chains in an aircraft brake system. As a result, this causes fatigue and various complex natural damages. Due to the complex surface conditions and the coexistence damages, it is extremely difficult to diagnose microcracks by using conventional thermography inspection methods. In this paper, the thermal pattern contrast method is proposed for weak thermal signal detection using eddy current pulsed thermography. In this process, the extraction and subsequent separation differentiate a maximum of the thermal spatial-transient pattern between defect and nondefect areas. Specifically, a successive optical flow is established to conduct a projection of the thermal diffusion. This directly gains the benefits of capturing the thermal propagation characteristics. It enables us to build the motion context connected between the local and the global thermal spatial pattern. Principal component analysis is constructed to further mine the spatial-transient patterns to enhance the detectability and sensitivity in microcrack detection. Finally, experimental studies have been conducted on an artificial crack in a steel sample and on natural fatigue cracks in aircraft brake components in order to validate the proposed method.

Thermographic inspection has been widely applied to non-destructive testing and evaluation with the capabilities of rapid, contactless, and large surface area detection. Image segmentation is considered essential for identifying and sizing defects. To attain a high-level performance, specific physics-based models that describe defects generation and enable the precise extraction of target region are of crucial importance. In this paper, an effective genetic first-order statistical image segmentation algorithm is proposed for quantitative crack detection. The proposed method automatically extracts valuable spatial-temporal patterns from unsupervised feature extraction algorithm and avoids a range of issues associated with human intervention in laborious manual selection of specific thermal video frames for processing. An internal genetic functionality is built into the proposed algorithm to automatically control the segmentation threshold to render enhanced accuracy in sizing the cracks. Eddy current pulsed thermography will be implemented as a platform to demonstrate surface crack detection. Experimental tests and comparisons have been conducted to verify the efficacy of the proposed method. In addition, a global quantitative assessment index F-score has been adopted to objectively evaluate the performance of different segmentation algorithms.

An unsupervised machine learning algorithm based on nonnegative matrix factor Two-dimensional deconvolution (NMF2D) with approximated optimum model order is proposed. The proposed algorithm adapted under the hybrid framework that combines the generalized EM algorithm with multiplicative update. As the number of parameters in the NMF2D grows exponentially the number of frequency basis increases linearly, the issues of model-order fitness, initialization, and parameters estimation become ever more critical. This paper proposes a variational Bayesian method to optimize the number of components in the NMF2D by using the Gamma-Exponential process as the observation-latent model. In addition, it is shown that the proposed Gamma-Exponential process can be used to initialize the NMF2D parameters. Finally, the paper investigates the issue and advantages of using different window length. Experimental results for the synthetic convolutive mixtures and live recordings verify the competence of the proposed algorithm.

The estimation of nutrient content of plants is considerably important in agricultural practices, especially in enabling the application of precision farming. A plethora of methods has been used to estimate nitrogen amount in plants, including the utilization of computer vision. However, most of the image-based nitrogen estimation methods are conducted in controlled environments. These methods are not so practical, time consuming, and require many equipment. Therefore, there is a crucial need to develop a method to estimate nitrogen content of plants based on leaves images captured on field. It is a very challenging task since the intensity of sunlight is always changing and this leads to an inconsistent image capturing problem. In this paper, we develop a low-cost, simple, and accurate approach image-based nitrogen amount estimation. Plant images are captured directly under sunlight by using a conventional digital camera and are subject to a variation in lighting conditions. We propose a color constancy method using neural networks fusion and a genetic algorithm to normalize various plant images due to different sunlight intensities. A Macbeth color checker is utilized as the reference to normalize the color of the images. We also develop a combination of neural networks using a committee machine to estimate the nitrogen content in wheat leaves. Twelve statistical RGB color features are used as the input parameters for the nutrient estimation. The obtained result shows considerable better performance than the conventional gray-world and scale-by-max approaches, as well as linear model and single neural network methods. Finally, we show that our nutrient estimation approach is superior to the commonly used soil-plant analysis development meter based prediction.

Traditional iris segmentation methods give good results when the iris images are taken under ideal imaging conditions. However, the segmentation accuracy of an iris recognition system significantly influences its performance especially in nonideal iris images. This paper proposes a novel segmentation method for nonideal iris images. Two algorithms are proposed for pupil segmentation in iris images which are captured under visible and near infrared light. Then, a fusion of an expanding and a shrinking active contour is developed for iris segmentation by integrating a new pressure force to the active contour model. Thereafter, a noncircular iris normalization scheme is adopted to effectively unwrap the segmented iris. In addition, a novel method for closed eye detection is proposed. The proposed scheme is robust in finding the exact iris boundary and isolating the eyelids of the iris images. Experimental results on CASIA V4.0, MMU2, UBIRIS V1, and UBIRIS V2 iris databases indicate a high level of accuracy using the proposed technique. Moreover, the comparison results with the state-of-the-art iris segmentation algorithms revealed considerable improvement in segmentation accuracy and recognition performance while being computationally more efficient.

Sclera blood veins have been investigated recently as a biometric trait which can be used in a recognition system. The sclera is the white and opaque outer protective part of the eye. This part of the eye has visible blood veins which are randomly distributed. This feature makes these blood veins a promising factor for eye recognition. The sclera has an advantage in that it can be captured using a visible-wavelength camera. Therefore, applications which may involve the sclera are wide ranging. The contribution of this paper is the design of a robust sclera recognition system with high accuracy. The system comprises of new sclera segmentation and occluded eye detection methods. We also propose an efficient method for vessel enhancement, extraction, and binarization. In the feature extraction and matching process stages, we additionally develop an efficient method, that is, orientation, scale, illumination, and deformation invariant. The obtained results using UBIRIS.v1 and UTIRIS databases show an advantage in terms of segmentation accuracy and computational complexity compared with state-of-the-art methods due to Thomas, Oh, Zhou, and Das.

Smart Home is a modern home that allows residents to have high-level comfort with effective use of electricity. These objectives can be achieved by applying suitable and promising optimization algorithms and techniques. This paper presents a demand side management strategy which was integrated into the existing Home Energy Management System (HEMS). Home energy management system is a Multi-Agent System (MAS) based decentralized architecture proposed by the authors. This intelligent energy management system was developed on an IEEE FIPA (Foundation for Intelligent Physical Agents) compliant multi-agent platform. This enables agents to communicate, interact and negotiate with energy sources and devices of the smart home to provide the most efficient energy usage and minimize the cost of electricity bills. This also results some peak load shaving of the power distribution system of the smart home. Simulation studies show the potential of proposed multi-agent system technique together with the demand side management strategy to provide the optimum solution for smart home energy management.

This paper proposes an unsupervised method for diagnosing and monitoring defects in inductive thermography imaging system. The proposed method is fully automated and does not require manual selection from the user of the specific thermal frame images for defect diagnosis. The core of the method is a hybrid of physics-based inductive thermal mechanism with signal processing-based pattern extraction algorithm using sparse greedy-based principal component analysis (SGPCA). An internal functionality is built into the proposed algorithm to control the sparsity of SGPCA and to render better accuracy in sizing the defects. The proposed method is demonstrated on automatically diagnosing the defects on metals and the accuracy of sizing the defects. Experimental tests and comparisons with other methods have been conducted to verify the efficacy of the proposed method. Very promising results have been obtained where the performance of the proposed method is very near to human perception.

This paper proposes a new system for the unsupervised diagnostic and monitoring of defects in waveguide imaging. The proposed method is automatic and does not require manual selection of specific frequencies for defect diagnostics. The core of the method is a computational intelligent machine learning algorithm based on sparse non-negative matrix factorization. An internal functionality is built into the machine learning algorithm to adaptively learn and control the sparsity of the factorization, and to render better accuracy in detecting defects. This is achieved by using Bayesian statistics methodology. The proposed method is demonstrated on automatic detection of defect in metals. In addition, we show that the extraction of the spectrum signature corresponding to the defect is significantly more efficient with the proposed optimal sparsity, which subsequently led to better detection performance. Experimental tests and comparisons with other sparse factorization methods have been conducted to verify the efficacy of the proposed method.

We present a new approach for solving the single channel speech separation with the aid of an user-generated exemplar source that is recorded from a microphone. Our method deviates from the conventional model-based methods, which highly rely on speaker dependent training data. We readdress the problem by offering a new approach based on utterance dependent patterns extracted from the user-generated exemplar source. Our proposed approach is less restrictive, and does not require speaker dependent information and yet exceeds the performance of conventional model-based separation methods in separating male and male speech mixtures. We combine general speaker-independent (SI) features with specifically generated utterance-dependent (UD) features in a joint probability model. The UD features are initially extracted from the user-generated exemplar source and represented as statistical estimates. These estimates are calibrated based on information extracted from the mixture source to statistically represent the target source. The UD probability model is subsequently generated to target problems of ambiguity and to offer better cues for separation. The proposed algorithm is tested and compared with recent method using the GRID database and the Mocha-TIMIT database.

A novel unsupervised machine learning algorithm for single channel source separation is presented. The proposed method is based on nonnegative matrix factorization, which is optimized under the framework of maximum a posteriori probability and Itakura-Saito divergence. The method enables a generalized criterion for variable sparseness to be imposed onto the solution and prior information to be explicitly incorporated through the basis vectors. In addition, the method is scale invariant where both low and high energy components of a signal are treated with equal importance. The proposed algorithm is a more complete and efficient approach for matrix factorization of signals that exhibit temporal dependency of the frequency patterns. Experimental tests have been conducted and compared with other algorithms to verify the efficiency of the proposed method.

Developing audio processing tools for extracting social-audio features are just as important as conscious content for determining human behavior. Psychologists speculate these features may have evolved as a way to establish hierarchy and group cohesion because they function as a subconscious discussion about relationships, resources, risks, and rewards. In this paper, we present the design, implementation, and deployment of a wearable computing platform capable of automatically extracting and analyzing social-audio signals. Unlike conventional research that concentrates on data which have been recorded under constrained conditions, our data were recorded in completely natural and unpredictable situations. In particular, we benchmarked a set of integrated algorithms (sound speech detection and classification, sound level meter calculation, voice and nonvoice segmentation, speaker segmentation, and prediction) to obtain speech and environmental sound social-audio signals using an in-house built wearable device. In addition, we derive a novel method that incorporates the recently published audio feature extraction technique based on power normalized cepstral coefficient and gap statistics for speaker segmentation and prediction. The performance of the proposed integrated platform is robust to natural and unpredictable situations. Experiments show that the method has successfully segmented natural speech with 89.6% accuracy.

The authors of the paper "Wearable Audio Monitoring: Content-Based Processing Methodology and Implementation" (ibid., vol. 44, no. 2, pp. 222-233) would like to acknowledge that the research reported was funded by the UK Medical Research Council grant "A monitoring device to objectively assess functional/psychosocial impairment in older-age adults with major depression" (G1001828/1).

Unsupervised extraction of focused regions from images with low depth-of-field (DOF) is a problem without an efficient solution yet. In this paper, we propose an efficient unsupervised segmentation solution for this problem. The proposed approach which is based on ensemble clustering and graph-cut modeling aims to extract meaningful focused regions from a given image at two stages. In the first stage, a novel two-level based ensemble clustering technique is developed to classify image blocks into three constituent classes. As a result, object and background blocks are extracted. By considering certain pixels of object and background blocks as seeds, a constraint is provided for the next stage of the approach. In stage two, a minimal graph cuts is constructed by utilizing the max-flow method and using object and background seeds. Experimental results demonstrate that the proposed approach achieves an average F-measure of 91.7% and is computationally up to 2 times faster than existing unsupervised approaches.

A novel single-channel blind source separation (SCBSS) algorithm is presented. The proposed algorithm yields at least three benefits of the SCBSS solution: 1) resemblance of a stereo signal concept given by one microphone; 2) independent of initialization and a priori knowledge of the sources; and 3) it does not require iterative optimization. The separation process consists of two steps: 1) estimation of source characteristics, where the source signals are modeled by the autoregressive process and 2) construction of masks using only the single-channel mixture. A new pseudo-stereo mixture is formulated by weighting and time-shifting the original single-channel mixture. This creates an artificial mixing system whose parameters will be estimated through our proposed weighted complex 2-D histogram. In this paper, we derive the separability of the proposed mixture model. Conditions required for unique mask construction based on maximum likelihood are also identified. Finally, experimental testing on both synthetic and real-audio sources is conducted to verify that the proposed algorithm yields superior performance and is computationally very fast compared with existing methods.

Automatic segmentation of images with low depth of field (DOF) plays an important role in content-based multimedia applications. The proposed approach aims to separate the important objects (i.e., interest regions) of a given image from its defocused background in two stages. In stage one, image blocks are classified into object and background blocks using a novel cluster ensemble algorithm. By indicating the certain pixels (seeds) of the object and background blocks, a hard constraint is provided for the next stage of the approach. In stage two, a minimal graph cut is constructed using object and background seeds, which is based on the max-flow method. Experimental results for a wide range of busy-texture (i.e., noisy) and smooth regions demonstrate that the proposed approach provides better segmentation performance at higher speed compared with the state-of-the-art approaches.

A novel approach for adaptive regularization of 2-D nonnegative matrix factorization is presented. The proposed matrix factorization is developed under the framework of maximum a posteriori probability and is adaptively fine-tuned using the variational approach. The method enables: (1) a generalized criterion for variable sparseness to be imposed onto the solution; and (2) prior information to be explicitly incorporated into the basis features. The method is computationally efficient and has been demonstrated on two applications, that is, extracting features from image and separating single channel source mixture. In addition, it is shown that the basis features of an information-bearing matrix can be extracted more efficiently using the proposed regularized priors. Experimental tests have been rigorously conducted to verify the efficacy of the proposed method.

A novel approach to solve the single-channel source separation (SCSS) problem is presented. Most existing supervised SCSS methods resort exclusively to the independence waveform criteria as exemplified by training the prior information before the separation process. This poses a significant limiting factor to the applicability of these methods to real problem. Our proposed method does not require training knowledge for separating the mixture and it is based on decomposing the mixture into a series of oscillatory components termed as the intrinsic mode functions (IMFs). We show, in this paper, that the IMFs have several desirable properties unique to SCSS problem and how these properties can be advantaged to relax the constraints posed by the problem. In addition, we have derived a novel sparse non-negative matrix factorization to estimate the spectral bases and temporal codes of the sources. The proposed algorithm is a more complete and efficient approach to matrix factorization where a generalized criterion for variable sparseness is imposed onto the solution. Experimental testing has been conducted to show that the proposed method gives superior performance over other existing approaches.

This paper reviews the state-of-the art of modelling, simulation and visualisation and reviews the recent development of modelling, simulation and visualisation software for Nondestructive Evaluation (NDE). Simulation and visualisation can assist in the design and development of electromagnetic sensing and imaging techniques and systems for nondestructive testing, feature extraction and inverse problems for quantitative nondestructive evaluation. After reviewing the state-of-the art of electromagnetic modelling and simulation, case studies from electromagnetic NDE research and development for eddy current distribution and thermography are discussed.

The quality of service (QoS) support for multimedia communications faces a big challenge in a fading wireless network. On one hand, conventional automatic repeat request (ARQ) schemes are not effective for small-scale fading channels with correlated errors due to consecutive retransmission failures. On the other hand, large-scale fading due to propagation loss or shadowing severely limits transmission range. A novel differentiated cooperative medium access control (MAC) protocol, called DC-MAC, is proposed to enhance the QoS support for multimedia communications while supporting service differentiation based on the IEEE 802.11e architecture. By enabling cooperative ARQ, the retransmission is initiated from an appropriate transmission queue of an appropriate relay node instead of the original source. Since unnecessary and useless retransmissions may intensify the node contention and degrade the system performance contrarily, a novel negative acknowledgement feedback mechanism is introduced for loss distinguishing and channel estimation such that cooperative retransmission will be employed only when necessary and only by competent nodes. Extensive simulations are conducted on the OPNET platform to analyse the performances of DC-MAC under both small-scale and large-scale fading. Simulation results show that the proposed scheme significantly improves the performances of both multimedia applications and best-effort data applications in terms of throughput, delay and coverage with moderate user contention.

In this paper, a novel solution is developed to solve the problem of separating noisy and post-nonlinearly distorted mixture. In the proposed work, the source signals are nonstationary and temporally correlated. A generative model based on hidden Markov model (HMM) is derived to track the nonstationarity of the source signal while the source signal itself is modeled by temporally correlated generalized Gaussian distribution (GGD) model. The maximum likelihood (ML) approach is developed to estimate the parameters of the proposed model by using the expectation maximization (EM) algorithm and the source signals are estimated by maximum a posteriori (MAP) approach. The strength of the proposed approach lies in the tracking of the nonstationarity of the source signal by HMM and the temporal correlation by the autoregressive (AR) source model. This has resulted in high performance accuracy, fast convergence and efficient implementation of the estimation algorithm. Simulations have been investigated to verify the effectiveness of the proposed algorithm and the results have shown significant improvement has been obtained when compared with nonlinear algorithm without using HMM.

Conventional biometric systems use the original biometrics for authentication, which exposes the users' identities to a risk of being compromised. The randomized Radon signature (RRS) is a cancellable biometric technique that protects face biometrics during authentication by using Radon transform and random projection. The extracted Radon signatures are generated from the parametric domain of the face and then projected into a random multi-space of a uniform distribution. The generated RRS templates are nonreversible, suitable for image-based statistical face classifiers and reissueable. In this paper the fisherface algorithm is used to conduct a comparison between the original face images and the RRS templates. Results have shown a dramatic 53.37% enhancement in the genuine and impostor distributions separation that leads to a 31.34% reduction in the equal error rate.

Biometrics has become a strong candidate to replace traditional authentication systems however biometric data in itself is vulnerable and requires protection. This paper presents a new method to protect face biometric data using one-way transformation in which original face images cannot be retrieved. The secure and reissueable templates are generated by utilizing the Radon transformed signatures of the face biometric and a multi-space random projection. Using an image-based statistical algorithm, authentication is conducted on the transformed templates without the need to reverse them back. Genuine and impostor distributions separation was also improved by 13.82% leading to a 41.35% reduction in the equal error rate.

The quality of service (QoS) support in wireless local area networks (WLANs) faces a big challenge due to time-correlated fading channel. Conventional automatic repeat reQuest (ARQ) schemes based on time diversity may result in consecutive retransmission failures degrading QoS severely. This paper proposes a novel Differentiated cooperative medium access control (MAC) protocol, called DC-MAC, to enhance QoS in WLANs based on the IEEE 802.11e architecture. The retransmission is initiated from an appropriate transmission queue of an appropriate relay node instead of the original source to exploit spatial diversity. A novel negative acknowledgement (NAK) feedback mechanism is introduced for loss distinguishing and channel estimation such that cooperative retransmission will be employed only when necessary and only by competent nodes. Simulations conducted on the OPNET platform show that the proposed scheme significantly improves the performances of both multimedia applications and data applications in terms of throughput and delay while supporting service differentiation.

Cooperative communication is becoming a promising technology for wireless networks by exploiting multipath fading instead of mitigating its impact. To integrate cooperative diversity into practical systems, efficient protocols are needed across the entire protocol stack. This paper presents a Cooperative Relay-Based Auto-Rate MAC protocol (CRBAR) to enhance the multi-rate capability along a long link. By leveraging the broadcast nature of the wireless medium and spatial diversity, a low-rate hop can be replaced by two high-rate hops via adaptive MAC-layer cooperation. To adapt to dynamical channel variation and network topology, the relay candidates adaptively select themselves as the relay nodes and determine the relay scheme and transmission rates based on the instantaneous channel measurements. System-level simulation study shows that CRBAR significantly outperforms traditional rate adaptation schemes in realistic scenarios.

A novel hybrid multiuser detection scheme that jointly uses linear and nonlinear interference suppression techniques is developed for high-speed direct-sequence code-division multiple-access communications in multipath frequency-selective fading channels. The detector detects signals in a symbol-by-symbol style. Conventional decorrelating detectors suffer from the noise enhancement problem, which becomes more serious for dispersive multipath channels. The proposed detector uses interference cancellation technology to reduce the rank of the expanded signal subspace and hence it preserves the advantages of the expanded decorrelating detector in terms of complete multiple access interference and intersymbol interference suppression and meanwhile avoids its disadvantage in terms of noise enhancement. Computer simulation shows clear superiority of the new detector to other existing methods.

In this paper, a novel solution is developed to solve blind source separation of postnonlinear convolutive mixtures. The proposed model extends the conventional linear instantaneous mixture model to include both convolutive mixing and postnonlinear distortion. The maximum-likelihood (ML) approach solution based on the expectation-maximization (EM) algorithm is developed to estimate the source signals and the parameters in the proposed nonlinear model. In the proposed solution, the sufficient statistics associated with the source signals are estimated in the E-step, while the model parameters are optimized through these statistics in the M-step. However, the complication resulted from the postnonlinear function associated with the mixture renders these statistics difficult to be formulated in a closed form and hence causes intractability in the parameter optimization. A computationally efficient algorithm is proposed which uses the extended Kalman smoother (EKS) to facilitate the E-step tractable and a set of self-updated polynomials is used as the nonlinearity estimator to facilitate closed form estimations of the parameters in the M-step. The theoretical foundation of the proposed solution has been rigorously developed and discussed in details. Both simulations and recorded speech signals have been carried out to verify the success and efficacy of the proposed algorithm. Remarkable improvement has been obtained when compared with the existing algorithms.

A novel polynomial-based neural network is proposed for nonlinear blind source separation. We focus our research on a recently presented mono-nonlinearity mixture where a linear mixing matrix is slotted into two mutually inverse nonlinearities. In this paper, we generalize the mono-nonlinearity mixing system to the situation where different nonlinearities are applied to the source signals. The theory of series reversion is merged with the neural network demixer to perform two layers of mutually inverse nonlinearities. The corresponding parameter learning algorithm for the proposed polynomial-based neural network demixer is also presented. Simulations have been carried out to verify the efficacy of the proposed approach. We demonstrate that the proposed network can successfully recover the original source signals in a blind mode under nonlinear mixing conditions

The post-nonlinear undercomplete blind signal separation problem is solved by a Bayesian approach in this paper. The proposed algorithm applies the generalized Gaussian model to approximate the prior distribution probability and a maximum a posteriori (MAP) based learning algorithm to estimate the source signals, mixing matrix and the nonlinearity of the mixing process. The mixing nonlinearity is modeled by a multilayer perceptron (MLP) neural network. In our proposed algorithm, the source signals, mixing matrix and the parameters of the MLP are iteratively updated in an alternate manner until they converges to a fixed value. The noise variance is regarded as the hyperparameter which is estimated in a closed form. Simulations based on real audio have been carried out to investigate the efficacy of the proposed algorithm. A performance gain of over 125% has been achieved when compared to linear approach

Linear independent component analysis (ICA) played an important role in the development of various signal processing techniques due to the inherent simplicity. However, the assumption of linear mixture is always violated in real life, which narrows down its applications. In this paper, the problem of nonlinear independent component analysis is considered. Based on a new type of nonlinear mixing model, we propose a two-stage series-based approach to recover the original source signals. The two-stage series-based algorithm offers significant advantages in terms of reduced computational complexity and better learning dynamics of the trajectory. Simulations have also been carried out to verify the efficacy of the proposed method

In this paper, a method of two dimensional Fisher principal component analysis (2D-FPCA) in the two dimensional principal component analysis (2DPCA) transformed space is analyzed and its nature is revealed, i.e., 2D-FPCA is equivalent to 2DPCA plus two dimensional Fisher linear discriminant analysis (2DFLD). Based on this result, a more transparent 2D FPCA algorithm is developed. That is, 2DPCA is performed first and then 2DFLD is used for the second feature extraction in the 2DPCA transformed space. Since 2D FPCA is based on the 2D image matrices, the vectorization of the image is not required. Thus, 2D FPCA optimizes the evaluation of the image matrices, the between and within matrices, by transforming them into a smaller 2DPCA space. In the linear discriminant analysis (LDA) based face recognition techniques, image representation and recognition is statistically dependent on the evaluation of the between and within matrices. This leads to the following benefits; the proposed 2D-FPCA yields greater recognition accuracy while reduces the overall computational complexity. Finally, the effectiveness of the proposed algorithm is verified using the ORL database as a benchmark. The new algorithm achieves a recognition rate of 95.50% compared to the recognition rate of 90.00% for the Fisherface method

This paper presents an iterative (turbo) channel estimation and turbo equalization when orthogonal frequency division multiplexing (OFDM) has been concatenated to space-time block code (STBC) over time-varying multipath Rayleigh fading channels. Based on the Kalman filter channel estimator, softinput- soft-output (SISO) minimum mean square error (MMSE) equalizer and a maximum posterior probability (MAP) decoder, the channel estimate can be improved for each iteration by using soft information feedback from the decoder. Computer simulations show good estimation and tracking by the Kalman filter with only a few training (pilot) symbols even when signals experience deep fading. It is observed from simulations that the overall system performance can be improved by 4dB after 14 iterations compared to non-iterative channel estimation and equalization. The proposed method is shown to obtain fast channel estimation and low complexity turbo equalization which are requirements in time-varying fading channels.

In this letter, a new type of nonlinear mixture is derived and developed into a multinonlinearity constrained mixing model. The proposed signal separation solution integrates the Theory of Series Reversion with a polynomial neural network whereby the hidden neurons are spanned by a set of mutually reversed activation functions. Simulations have been undertaken to support the theory of the proposed scheme and the results indicate promising performance.

Conventional assumptions of square mixing matrix and negligible noise adopted in blind signal separation do not always correspond with real applications. Signal detection from a small number of sensors is often required in signal and image modeling and biomedical applications. The paper proposes a new algorithm to estimate accurately signals from underdetermined mixtures with fewer restrictions and assumptions compared with existing techniques. The strength of this algorithm is that it does not adopt the conventional assumptions on the mixing, signals and noise. The algorithm is capable of separating orthogonal and non-orthogonal mixtures of both sparse and non-sparse signals with additional Gaussian or nonGaussian noise. This algorithm is also applicable to separating time-varying and instantaneous mixtures. Simulation results demonstrate the efficacy of the proposed algorithm for separation of time-varying mixtures in the presence of noise.

Tracking human motion in monocular video is a challenging problem in computer vision. It has found a wide range of applications, such as visual surveillance, virtual reality, sports science, etc. This project aims to develop a model-based human motion analysis system that can track human movement in a monocular image sequence with minimum constraint. No markers or sensors are attached to the subject. Given a video clip, the first step is to fit the 3D human model manually to the subject in the first frame of the video. Then background subtraction is used to extract the human silhouette. We propose the silhouette chamfer as the main matching feature. A chamfer distance measure is carried out on the extracted subject silhouette. The silhouette chamfer contains both the chamfer distance and region information. Finally, we use a discrete Kalman filter to predict the pose of the subject in each image frame. The updating step uses Broydent's method to optimize the predicted pose to fit the person's silhouette by using the cost function. We use the gait database SOTON to test our system. The image sequences contain human walking in both indoor and outdoor environments. The motion tracking results demonstrate that our system has an encouraging performance.

In multiple input multiple output systems (MIMO), the increase in the number of channel parameters, renders less reliable conventional training based schemes. For a reasonable performance, conventional training based methods require a larger number of training symbols which reduces system bandwidth efficiency. In this paper, a semiblind channel estimation and tracking technique is proposed for concatenated turbo codes (TC) with space-time and frequency block codes (STFBC) over time-varying Rayleigh fading channels. The paper describes an effective approach to reduce the number of training symbols to half by implementing the Kalman filter (KF) with the constant modulus algorithm (CMA). Computer simulations show good estimation and tracking by the Kalman filter and CMA with only few training symbols (less than 7% of bandwidth efficiency loss) even when signals experience deep fades. The proposed method is shown to obtain fast channel estimation and tracking which is a requirement in fast fading channels.

A space-time nonlinear receiver for a DS-CDMA synchronous channel is proposed for jointly mitigating impulsive noise and multiple access interference. The proposed scheme combines linear decorrelators and antenna arrays with nonlinear front ends based on the affinity matched myriad nonlinear filter. This myriad (log Cauchy) filter is incorporated to combat the impulsive noise and prevent it from entering the system. The proposed detector also incorporates a structure for estimating the impulsive noise parameter in order to update a filter parameter called the nonlinearity parameter. Monte Carlo simulation results of the proposed space-time receiver are presented to justify the relative merits of nonlinear signal processing in the space-time domain.