In Internet of Things (IoT) applications, privacy-preserving concerns restrict direct data sharing across distributed devices. Moreover, disparities in device performance, operational environments and sensing capabilities lead to highly imbalanced data distributions across clients, undermining model generalization and limiting the adaptability to diverse data distributions. While Federated Learning (FL) enables collaborative training without raw data exchange, it remains vulnerable to imbalanced data distributions across clients, often resulting in global model bias and poor generalization. Existing methods mitigate this by sharing partial statistical representations, yet fail to comprehensively capture complex data distributions. To address this issue, we propose Uncertainty-guided Gaussian Representation Sharing (UGRS), a novel FL method designed to capture more comprehensive distributional characteristics. UGRS shares global Gaussian representations to guide local training on clients and dynamically refines these representations based on sample uncertainty. Gaussian representations inherently encode holistic data distribution features, making them more effective than partial statistical features in reflecting the complexity of the entire data distribution. Additionally, UGRS constructs compact Gaussian representations by projecting high-dimensional features into a low-dimensional space—this not only reduces data transmission overhead but also minimizes privacy leakage risks by limiting exposure of sensitive high-dimensional information. Experiments on MNIST, CIFAR-10, CIFAR-100, Tiny-ImageNet, and a custom IoT dataset with client-specific imbalanced distributions demonstrate that the UGRS consistently outperforms state-of-the-art methods with an average accuracy improvement of 0.87%–5.72% for both global and local models.

With the rapid development of Internet of Things (IoT), accurate and robust two-dimensional direction of arrival (2-D DOA) estimation has become more and more crucial. This paper addresses the significant challenge of 2-D DOA estimation with high degrees of freedom (DOF) under impulsive noise in practical IoT systems. Firstly, an augmented extended coprime parallel array is proposed by integrating an extended coprime array (ECA) with an augmented unfolded coprime array. The proposed design exploits non-circular signal characteristics, achieving an enhancement in DOF. Furthermore, a generalized versoria function (GVF) is introduced to suppress impulsive noise, and a GVF-based covariance matrix is developed to achieve robust DOA estimation. Simulation results confirm the proposed method achieves higher estimation accuracy and DOF under impulsive noise, making it suitable for reliable IoT systems.

Multiple-input multiple-output wireless sensing systems achieve high-resolution detection through spatial diversity. However, they suffer from reliability degradation under low signal-to-noise ratio (SNR) conditions. Conventional matrix-based methods may discard critical target information embedded in inter-dimensional signal correlations due to dimension-reduction flattening operations. To overcome these limitations, this paper proposes a tensorial target detection (TTD) framework combining noise reduction and enhanced detection specifically designed for high-dimensional processing. Firstly, we propose a two-stage tensorial noise reduction (TNR) method based on the minimum mean square error criterion and the alternating least square iteration strategy to remove noise in high-order signal space. We further identify the sub-optimal performance caused by inter-dimensional noise coupling at the first stage of TNR, and resolve the issue via rank-constrained optimization for noise-target subspace separation at the second stage of TNR. Then, we develop an augmented tensorial detector based on cross-shaped spatial partitioning (CSP) to enhance detection performance, which jointly optimizes detection thresholds by adaptively refining noise estimation. Finally, field measurements confirm the TTD framework's operational validity, while in simulation the TNR method achieves 5 dB SNR improvement over 2D method, and the CSP-based detector delivers a 21% enhancement in detection probability over conventional approaches.

Voice control has become an indispensable interaction method in smart devices. Compared to traditional microphones, mm Wave radar offers a promising solution for speech sensing in noisy environments. However, most current research relies on single-view information, such as vocal cord vibrations or lip movements, to classify speech, which overlooks important details like timbre, speech rate, and intonation, limiting the application of speech sensing. To address these issues, we develop a high-quality speech sensing method based on mm Wave radar, named SpDiff. This method accurately localizes the vocalizing target and, based on the human vocal mechanism, extracts multi-view speech features according to the movement characteristics of the vocal cords, lips, and face. Additionally, to generate high-quality speech signals, we design a conditional latent diffusion model (CLDM), which uses multi-view radar information as conditional guidance, accurately capturing the complex mapping relationships between radar and speech signal distributions. To evaluate the SpDiff method, we build a mmWave system using IWR1443Boost and recruit 14 volunteers to construct a dataset. Experimental results show that SpDiff achieves high standards in speech sensing, with the generated speech directly input into existing recognition models, achieving an average character and word error rate (CER/WER) of only 2.33% and 3.05%.

Device-free wireless sensing (DFWS) has gained significant attention due to its high accuracy and privacy-preserving capabilities. DFWS systems work by analyzing the influence pattern of targets on the surrounding wireless signals. However, changes in the sensing scenario can alter signal propagation patterns, causing deep learning models to lose effectiveness in cross-scenario applications. To address this problem, we analyze the information content of different samples from a free-energy view, and provide a new idea to guide the alignment of target scenario samples with source scenario samples based on free-energy. We find that free-energy can measure the degree of scenario knowledge contribution of the samples. Based on this observation, we first perform unsupervised coarse alignment by minimizing the free-energy deviation between scenarios. Next, we iteratively select a few number of high-free-energy samples near the decision boundary to fine-tune the network, achieving scenario fine alignment with a small labeling effort. Extensive experiments on two public datasets and one self-collected dataset show that our proposed method achieves high accuracy in cross-scenario human activity and gesture recognition tasks.

Due to the uniqueness and stability of radio frequency fingerprints (RFF), radio frequency fingerprint identification (RFFI) is an important physical layer authentication method in the security of Internet of Things (IoT). However, existing deep-learning-based RFFI methods require a large number of labeled samples to achieve ideal performance. Besides, when the signal length changes, the network structure needs to be redesigned and the entire training process needs to be reconducted. To address this issue, we propose a few shot RFFI method on the basis of self-distillation with no labels (SDINO). Within it, a network termed DPformer is designed, which can adapt signals of varying lengths and is more lightweight. When the signal length changes, there is no need to retrain the network, and it is more lightweight. Simulation results show that, compared with existing methods, the proposed method achieves better recognition performance and more lightweight on LoRa dataset with 30 classes.

Visual trajectory estimation can endow autonomous underwater vehicles (AUVs) with environmental perception capabilities and has broad application prospects in the fields, such as oceanographic surveys, underwater construction, and marine ranching. However, due to featureless images and depth ambiguity issues caused by underwater multiple mediums environments, monocular visual trajectory estimation in complex underwater environments remains a challenging problem. In this article, we propose a monocular visual trajectory estimation method for AUVs, which can address the challenges of featureless and depth ambiguity by leveraging deep image representations and multiview geometry. Specifically, we design a bidirectional optical flow consistency scheme that selects sparse correspondences from monocular dense predictions to deal with the featureless images, and then achieve AUV trajectory estimation through epipolar constraints. Furthermore, we propose an iterative depth-consistent method, which solves the problem of depth ambiguity by aligning geometrically triangulated depths to the scale-consistent deep depths. We also develop a low-cost, agile, and portable AUV picking system with real-time trajectory estimation capabilities, and carry out extensive experiments in the Yellow Sea to test its performance. The experimental results demonstrate the effectiveness of the proposed method.

Unmanned aerial vehicle (UAV) assisted communication has emerged as a powerful technology for reliable and flexible emergency communications (e.g., earthquakes, hurricanes and floods), especially when the mobile infrastructure is seriously damaged. UAV assisted mobile edge computing (UAV-MEC) system can be deployed in the natural disaster area as communication relay or air mobile base stations to resume communication and provide computing resources for the users in disaster areas. However, the optimized resource allocation performance of UAV-MEC system can be further guaranteed with the human-fusion decision making. In this paper, we construct an emergency human-machine fusion UAV-MEC system consisting of multiple UAVs equipped with computing resources, and the human-machine decision makings are fused for UAV deployment. In order to solve the resource allocation problem of human-machine fusion UAV-MEC system, we establish an human-machine deep integration model for UAV-MEC system, and the UAVs are dispatched reasonably through human-machine fusion decision makings to maintain efficient communication in emergency communication areas. To minimize task latency and improve the computation efficiency in emergency human-machine fusion UAV-MEC system, we consider the number of dispatched UAVs, deployment plans, flight plans, and simultaneously optimize the task allocation scheme, priority order, and task offloading ratio. We propose a reinforcement learning framework combined with evolutionary algorithms, which is named as multi-agent Q-net enhanced cooperative genetic algorithm (MQCGA), for resource allocation of UAV. Based on neural network forecasts, the greedy rate during training processing can be dynamically controlled, and the learning ability of different agents can be strengthened. Simulation experiments are conducted to evaluate the proposed framework, and the results show that our proposed MQCGA algorithm is significantly superior to other algorithms in terms of latency and energy consumption. Note to Practitioners—With the development of MEC and the popularity of UAVs, the potential of UAV-assisted MEC draw much attention from industrial field. Considering the lack of communication capabilities in a certain area in an unexpected situation, UAVs can be quickly deployed to corresponding locations and provide computing services. In this paper, an UAV-MEC system that integrates human-machine decision-making for emergency communication situations, named human-machine fusion UAV-MEC system, is considered. The system divides the scene into regions, models users and UAVs, and provides detailed deployment schemes, maximizing the practicality and applicability of the scene. In order to improve the communication efficiency in the case of emergency communication, this paper proposes a new resource scheduling algorithm and adds human-machine decision-making to enable UAVs to continuously provide efficient services for a certain area. The experimental results provide practitioners with a theoretical basis, such as the task completion time, UAV energy consumption and computing resource scheduling. Applying the system to actual scenarios also requires two preconditions of the system, one is the information collected by the large UAV, and the other is the communication among the UAVs. These two preconditions facilitate the human-machine fusion UAV-MEC system deployment in practical applications.

Underwater image enhancement plays a pivotal role in addressing the challenges posed by the complex and dynamic underwater environment. While the previous research has conducted valuable explorations from a global enhancement perspective, underwater settings often exhibit multidistribution characteristics in both spatial frequency and illumination conditions that require specialized attention, that is, multiple spatial frequencies and lighting conditions coexist in the same image, making it difficult to achieve optimal enhancement using global mapping. To address this challenge, we propose a multidistribution aware network (MDA-Net) that leverages local frequencies and illumination characteristics of images for adaptive adjustment to balance the diverse visual enhancement requirements of local regions. Specifically, to address the challenge of multiple spatial frequency distributions, we explore the correlation among spatial frequency, receptive field, and image quality perception, and design a frequency-aware kernel selection convolution, which could adaptively select the size of convolutional kernels based on the frequency complexity of each region, so as to balance the requirements of noise reduction and color fidelity in different regions. Furthermore, to address the challenge of multiple illumination distributions, we leverage the inherent illumination characteristics of the image to generate a gamma transformation-based illumination balancer (GIB), whose neurons can comprehensively perceive global and local illumination through multiparameter correction representation, thereby guiding the focus of the enhancement work. Extensive experiments with the ablation analysis show the effectiveness of our proposed MDA-Net on four benchmark datasets: UFO-120, UIEB, UIEB-U60, and U45.

With the advancement of spectral imaging technology, hyperspectral image (HSI) resources have been rapidly expanding, making cross-dataset HSI classification a critical technique and an inevitable trend for large-scale Earth observation applications. However, most existing approaches transfer from one HSI to another in a supervised way, which limits their ability to leverage multi-source HSI information. Therefore, this paper proposes an unsupervised cross-dataset HSI classification method based on parallel adversarial domain adaptation (PADA), which learns and integrates task-relevant and domain-invariant knowledge from multi-source HSIs to classify a target HSI. Specifically, a source-and-target shared information mining module is designed to mine transferable knowledge from each source HSI. This module employs parallel adversarial learning between a spectral-spatial feature extractor and a task-relevant controller with a domain-invariant discriminator to learn task-relevant and domain-invariant features, thereby mitigating domain shift. Moreover, a source-to-target transferability learning module is proposed to evaluate the cross-dataset transferability of knowledge, which computes domain correlation score to evaluate inter-domain correlation and guide adaptive knowledge transfer, effectively suppressing redundant information and reducing negative transfer caused by low-correlated sources. Finally, a multi-source collaborative classification module is developed to accomplish cross-dataset knowledge transfer, which designs correlation-aware fusion strategy to produce the final classification result by integrating information from each source, ensuring balanced and robust decision-making. Extensive experiments conducted under challenging multi-source cross-dataset settings validate the classification performance and the domain extensibility of the proposed method, demonstrating its potential for large-scale Earth observation applications.

Ensuring resilient semantic segmentation under diverse outdoor conditions is vital for autonomous driving. However, nighttime segmentation remains underdeveloped compared to its daytime counterpart due to poor illumination and scarce annotated data, posing a significant challenge for night-scene understanding. Most current approaches mainly rely on domain adaptation technologies to transfer segmentation models trained on daytime scenes. However, the substantial distinctions between daytime and nighttime domains often hinder effective adaptation. To address this challenge, we leverage the implicit comprehensive information within nighttime data to enhance semantic segmentation through a semi-supervised approach. Specifically, we introduce a Semi-Supervised Comprehensive Learning (SSCL) approach, which is a unified, closed-loop learning architecture composed of three mutually reinforcing correction mechanisms: (1) unsupervised interactive correction to mitigate the risk of erroneous label propagation by leveraging complementary learning abilities; (2) unsupervised reinforcement correction, which enhances the model’s adaptability by promoting diverse learning on high-uncertainty regions through entropy-guided perturbation; (3) supervised standard correction to ensure alignment with known standards by anchoring the system to reference answers. SSCL is the first semi-supervised algorithm that jointly exploits structural diversity, uncertainty-aware supervision, and closed-loop correction to fully harness the latent potential of unlabeled nighttime data. Extensive experiments on NightCity, Dark Zurich and Nighttime Driving datasets demonstrate that SSCL achieves state-of-the-art performance in nighttime semantic segmentation.

Micro-mobility has become a vital means of transportation in recent years, however, it has also resulted in a rise in traffic incidents. Timely tracking and predicting riders' maneuvers hold the potential to ensure active protection and allow for sufficient time to avert accidents by issuing timely warnings and interventions. We contend that the rider's head dynamics can provide valuable information regarding their subsequent maneuvers. Riders' traveling habits, however diverse, not to mention the rapidly varying riding environment. The above factors contribute to significant disruptions in the data source, and various micro-mobility forms further exacerbate the issue. We accordingly present HeadMon+, which predicts the rider's subsequent maneuver by examining their head dynamics, and it can effectively adapt to various riding conditions and individuals. The system incorporates a deep learning framework with an advanced domain adversarial network. By single-time pre-training, HeadMon+ is capable of adapting to new data domains, including human subjects, and riding conditions for robust maneuver prediction. Based on our evaluation, we have found that the maneuver prediction of HeadMon+ has an overall precision of 94% with a prediction time gap of 4 seconds. HeadMon+'s low cost and rapid response capability make it easily deployed and then contribute to enhancing safe riding.

Device-Free mmWave Sensing (DFWS) could sense target state by analyzing how target activities influence the surrounding mmWave signals. It has emerged as a promising sensing technology. However, when employing DFWS indoors, specular reflection interference arises due to the specular reflectors. This interference often induces ghost targets, impacting the accurate estimation of the number and position of targets, resulting in degradation in sensing performance. To tackle this issue, we delve into the generation mechanism of specular reflection interference and analyze its multi-domain characteristics. Through exploration, we discern its temporal sparsity, spatial symmetry or collinearity, and frequency correlation characteristics, and propose four metrics to measure them, accordingly. Specifically, we propose a temporal characteristic quantitative evaluation metric based on identity matching, spatial symmetry and collinearity quantitative evaluation metrics based on geometric analysis, and a frequency correlation quantitative evaluation metric based on Doppler velocity correction, respectively. Based on these metrics, we design a novel Specular Reflection Interference Mitigation (SRIM) method and develop a robust SRIM-DFWS prototype system based on a 60 GHz mmWave radar to validate our proposed method. Experimental results demonstrate that our proposed method could achieve accurate and effective mitigation of specular reflection interference in device-free target tracking.

Device-free positioning (DFP) using mmWave signals is an emerging technique that could track a target without attaching any devices. It conducts position estimation by analyzing the influence of targets on their surrounding mmWave signals. With the widespread utilization of mmWave signals, DFP will have many potential applications in tracking pedestrians and robots in intelligent monitoring systems. State-of-the-art DFP work has already achieved excellent positioning performance when there is one target only, but when there are multiple targets, the time-varying target state, such as entering or leaving of the wireless coverage area and close interactions, makes it challenging to track every target. To solve these problems, in this paper, we propose a spatial-temporal analysis method to robustly track multiple targets based on the high precision mmWave point cloud information. Specifically, we propose a high precision spatial imaging strategy to construct fine-grained mmWave point cloud of the targets, design a spatial-temporal point cloud clustering method to determine the target state, and then leverage a gait based identity and trajectory association scheme and a particle filter to achieve robust identity-aware tracking. Extensive evaluations on a 77 GHz mmWave testbed have been conducted to demonstrate the effectiveness and robustness of our proposed schemes.

Empowered by the feature extraction ability of deep neural networks (DNNs), the DNN-based device-free wireless sensing (DFWS) could recognize human activity by analyzing the pattern information involved in the influenced wireless signals. However, labeling samples is time-consuming and labor-intensive because wireless signals are not human-interpretable. In practical applications, there are always few labeled samples, and how to realize high-performance DFWS with few labels becomes an urgent problem to solve. To tackle this problem, finding compact representative features for samples in an unsupervised manner is crucial. To this end, we design a contrastive learning framework to obtain features of unlabeled samples by maximizing the mutual information between features and the corresponding samples. The contrastive training process extracts features for the input samples by contrasting positive and negative sample pairs, thus strengthening the correlation between the features and the corresponding samples. The intuition behind our method is that mutual information measures the correlation between features and samples, and thus the maximum mutual information could capture informative and discriminative features. Our evaluation results on two publicly available data sets and one data set collected by ourselves show that our proposed method achieves satisfactory accuracy for both human activity and gesture recognition tasks with few labels.

As a promising solution to alleviate network congestion, mobile edge caching based on unmanned aerial vehicles (UAVs) has emerged and received intensive research interests, where users could download their desired contents from UAVs with much lower latency. As for the UAV-assisted edge caching, to improve the users’ Quality of Experience while reducing the cost on content updating, how to jointly design the trajectory and caching strategy for UAVs is critical. However, considering the dynamics and uncertainty on the traffic environment, as well as the mutual effect among different UAVs, such joint design is nontrivial. In this article, we propose a collaborative joint trajectory and caching scheme for UAV-assisted networks under the dynamic and uncertain traffic environment. Unlike most existing work relying on model-based or single-agent methods, we develop a multiagent deep reinforcement learning (MADRL) approach to obtain the solution, where the specific content demand model is not needed and each UAV would learn the best decision autonomously based on its local observations. It can achieve the adaptive cooperation among different UAVs, while optimizing the overall network performance. Moreover, standing from the perspective on swarm intelligence, we further develop a dynamic clustering federated learning framework on the MADRL algorithm. By performing parameter fusion, each UAV can improve the learning efficiency.

Human gesture recognition, due to its indispensable role in a myriad of emerging applications, has attracted close attention in the visual and wireless sensing community. The intrinsic characteristics of visual and wireless modalities are complementary to each other, e.g., wireless signals are robust to illumination changes and occluded conditions but suffer from low-spatial resolution, while visual signals have a high-spatial resolution but are vulnerable to scenario variations. Intuitively, integrating them has the potential chance to improve the overall discriminative ability. However, due to their different physical patterns and semantics, how to explore the relationship between the two modalities and leverage their complementary information to improve recognition performance still remains unsolved. In this article, we propose a complete representation projection method, which projects the signals from the heterogeneous modalities into a complete representation feature space by performing a bidirectional projection constraint. Furthermore, to relieve the low-projection efficiency problem caused by the heterogeneity of the two modality features, we propose an attention-based cross-modality interaction (ACMI) mechanism to perform implicit semantic feature alignment, thereby better capturing the complex dependencies between the two modalities and improving the feature projection efficiency. To evaluate the proposed method, we build a visual-radar cross-modality gesture recognition system and conduct extensive experiments. Experimental results demonstrate that the proposed approach not only performs favorably against vision-only and wireless-only solutions by a large margin, but also shows superiority over traditional fusion solutions.

Human gesture recognition provides great potentials in Human Computer Interaction (HCI), and the wireless or visual signals based technologies have been explored in their respective fields. The intrinsic characteristics of both modalities are complementary to each other, e.g. the wireless signal is robust to illumination changes and occluded conditions but suffers from low space resolution, while the visual signal has high space resolution but vulnerable to scenario variations. Intuitively, integrating the two modalities has potential chance to improve the overall discriminative power. However, existing multi-modal fusion methods could not fully exploit their complementarity to achieve accurate estimation, and also lack physical interpretability. In order to solve this issue, we introduce WiVi-GR: a Wireless-Visual joint representation based accurate Gesture Recognition system, which constructs a complete velocity representation to guarantee robust and accurate gesture recognition. Specifically, we analyze the complementarity of the two modalities in data dimension and spatial-temporal feature resolution, and propose an Interpretable Orthogonal Representation (IOR), which applies multi-channel coding to get image plane velocity, utilizes frequency domain analysis to get radial velocity, and aggregates both to achieve the complete representation of the dynamic pattern. Based on the IOR, we perform a data-level fusion with channel superposition convolutions to accomplish the accurate gesture recognition task. Experimental results show that the proposed WiVi-GR outperforms traditional multi-modal approaches by large margins, especially in small training sample set condition.

Unmanned aerial vehicles (UAVs) can be deployed in the areas where traditional network infrastructure is insufficient or absent because of flexibility and collaboration. The deployment of edge intelligence on UAVs in UAV-assisted mobile-edge networks significantly enhance data processing efficiency, which is a critical factor for applications requiring delay-sensitive data processing in the areas mentioned above. However, the limited energy capacity poses a challenge when running complex deployment algorithms. Therefore, lightweight network model is essential for deployment algorithms, as it significantly reduces energy and time consumption. In this article, we propose a cooperative framework-based knowledge distillation to compressed deep neural network (DNN). The subnetworks collaborate to train interactive node parameters, resulting in the optimal evolution of the student network. Then, we introduce a novel result-driven model training approach for simulation data sets. To further enhance efficiency and significantly reduce overall latency, we meticulously refine the internal architecture of the knowledge distillation algorithm. We incorporate a collaborative evolution mechanism into the core of the algorithm, utilizing multinetwork and subnetwork learning to facilitate knowledge transfer, and incorporate some optimization mechanisms into the framework. Finally, we perform a series of experiments to acquiredata sets and conduct algorithm simulation analysis to evaluate the proposed method. The results demonstrate that our work achieves good research results.

Device-free gesture recognition has attracted significant attention due to its potential applications in pervasive interaction. It enables gesture recognition in a device-free and contact-free manner by analyzing the influence pattern of human gestures on surrounding wireless signals, such as mmWave signals. Although remarkable progress has been achieved in this area, the recognition performance will degrade remarkably when gestures are conducted in different scenarios. In this paper, we leverage mmWave signals to design two robust trajectory features, i.e., the trajectory image and the trajectory time-sequence features, that are independent of the conducted scenarios to solve the aforementioned problems. Specifically, we employ the particle filter algorithm to construct the raw trajectory image utilizing range measurements, rotate and enhance the image to obtain the trajectory image feature suitable for recognition by leveraging a public handwriting font image data set as the training set. Additionally, we derive the range of the trajectory relative to a stable point as the trajectory time-sequence feature. With these trajectory features, we design a deep network to perform the gesture recognition task. To validate the effectiveness of the proposed methods, we conduct extensive experiments on a 77GHz mmWave testbed. The results indicate that the two proposed trajectory features are feasible for achieving scenario-independent gesture recognition.

Accidents are inevitable in the transportation systems; however, harnessing the big data generated from traffic accidents can significantly enhance the intelligence of the transportation system. Multiple unmanned aerial vehicles (multi-UAVs), owing to its flexibility and collaboration, can rapidly track the accidents and gather the acquire relevant data with reasonable deployment. Therefore, we propose an architecture utilizing multiple UAVs to track traffic accidents (targets) and collect relevant data. However, dynamic factors such as no-fly zones and reappearing targets in real-world traffic environments pose challenges to the rapid deployment of UAVs with current algorithms. In this paper, we design a model for deploying multiple UAVs, considering no-fly zones and dynamically reappearing targets. This model aims to optimize UAV deployment by minimizing both the flight distance of each UAV and the associated risk. Risk is defined as the urgency of processing accident scenes, which escalates with the elapsed time from the initial appearance of the targets to their processing. To address the joint UAV deployment problem, we first introduce an algorithm termed preprocessing and group-crossover nondominated sort genetic algorithm II (PGC-NSGAII). In PGC-NSGAII, we employ a group-based selection crossover (GBSC) method to enhance the algorithm’s search capability. This method segregates the initial population into two groups, selecting two individuals for crossover within each group. Secondly, building upon the framework of NSGAII, we incorporate a novel preprocessing component to enrich solution diversity. Thirdly, we develop a prediction method for dynamic environment parameters and a no-fly zone avoidance strategy for multi-UAV deployment. Finally, our experimental results demonstrate that PGC-NSGAII surpasses other existing methods in scenarios with varying numbers of UAVs or targets. Compared with state-of-the-art optimization methods, PGC-NSGAII proves to be more efficient in UAV deployment in dynamic environment.

Using mmWave radar to conduct gesture recognition is a promising solution for human-computer interaction. Although many studies have shown initial success, two-fold problems still remain unsolved, namely, the high-strength human activity interference and the difficulty in handling similar gestures. In light of these, we develop a robust mmWave radar based gesture recognition system, Rodar, to achieve accurate recognition of similar gestures under high-strength human activity interference, where a Multi-view De-interference Transformer (MvDeFormer) network is proposed. Specifically, to deal with the strong human activity interference, we design a DeFormer module to capture the useful gesture features by learning different patterns between gestures and interference, thereby reducing the impact of interference. Then, we develop a hierarchical multi-view fusion module to first extract the enhanced features within each view, and effectively fuse them across various views for final recognition. To evaluate the proposed Rodar system, we construct a dataset with seven similar gestures under three common human activity interference scenarios. Experimental results show that the accuracy can achieve up to 93.01%. The code implementations are available athttps://github.com/Xlab2024/MvDeFormer.

Device-free vital signs monitoring is an emerging technology that utilizes the unique influence of chest vibrations on surrounding wireless signals to achieve vital signs monitoring in a device-free and contact-free manner. Existing methods could achieve good monitoring performance when high-quality reflected signals can be obtained. However, in daily vital signs monitoring at home, the received reflected signals are often very weak due to factors such as obstruction and attenuation, resulting in a sharp decrease in the monitoring performance. To address the aforementioned challenges, in this paper, we develop a diversity-enhanced robust device-free vital signs monitoring system using mmWave signals. Specifically, inspired by the concept of diversity in the field of communications, we propose a diversity-enhanced wireless sensing strategy that comprehensively utilizes multi-dimensional physical layer resources, including antennas, chirps, and space, to improve the signal-to-noise ratio of vital signs. Additionally, inspired by cameras that achieve clear images by prolonging exposure time, we propose an accumulation-enhanced localization method to lock onto the chest of the human body in complex scenarios. Extensive experiments on a 60 GHz mmWave testbed demonstrate that our developed system could guarantee robust vital signs monitoring performance in various challenging scenarios, even at distances of up to 40 m.

Hyperspectral target detection (HTD) is an important issue in earth observation, with applications in both military and civilian domains. However, conventional representation-based detectors are hindered by the reliance on the unknown background dictionary, the limited ability to capture nonlinear representations using the linear mixing model (LMM), and the insufficient background-target recognition based on handcrafted priors. To address these problems, this paper proposes an interpretable representation network that intuitively realizes LMM for HTD, making nonlinear feature expression and physical interpretability compatible. Specifically, a subspace representation network is designed to separate the background and target components, where the background subspace can be adaptively learned. In addition, to further enhance the nonlinear representation and more accurately learn the coefficients, a lightweight multi-scale Transformer is proposed by modeling long-distance feature dependencies between channels. Furthermore, to supplement the depiction for target-background discrimination, a constrained energy minimization (CEM) loss is tailored by minimizing the output background energy and maximizing the target response. The effectiveness of the proposed method is demonstrated on four benchmark datasets, showing its superiority over state-of-the-art methods. The code for this work is available at https://github.com/shendb2022/HTD-IRN for reproducibility purposes.

Device-free wireless sensing (DFWS) has drawn lots of attention due to its potential application in the fields of human–computer interaction and smart home. Deep networks-based DFWS technique has achieved excellent sensing performance. However, network complexity limits its deployment on resource-limited sensing devices. A feasible way is to implement a simple network to accomplish the DFWS task. However, the sensing performance will drop dramatically due to its limited learning ability. In this article, to realize lightweight DFWS with acceptable performance, we propose an information-for-complexity strategy to promote the learning ability of the simple network. We leverage knowledge distillation framework to explore external information to augment the extrinsic learning ability, and utilize multiscale receptive fields to explore the internal information to augment the intrinsic learning ability. Extensive experiments on a 77 GHz mmWave testbed show that the performance degradation of the lightweight DFWS system is within 3%, while the complexity decreases remarkably.

Place recognition based on 3D point cloud can directly describe the scene of 3D world by acquiring 3D point clouds via LiDAR, which is robust with environmental changes. The core challenge locates at how to obtain compact and representative feature expression for positioning. In this paper, SparseARFM-SI framework is proposed to solve the problem of feature extraction caused by the large size difference of different objects and the difficulty of feature extraction based on the point cloud data collected by rotating LiDAR. SparseARFM-SI is mainly composed of four models: data representation model, sparse convolution model, ARFM model and NetVLAD model. Experiments show that SparseARFM-SI framework has good performance in USyd data set collected based on rotating Lidar.

Device-free gesture recognition (DFGR) is an emerging technique which could leverage the influence of human gestures on surrounding wireless signals to recognize gestures. It has gained widespread attention due to its promising prospect of empowering pervasive wireless devices with the sensing ability. Due to the inconsistency of the feature distribution in different scenarios, a well-trained DFGR system often fails to get satisfactory performance in cross-scenario conditions. Researchers have done valuable exploration on alleviating the feature distribution shift from a global distribution point of view. However, global feature distribution alignment could not solve the feature distribution shift problem completely. In this article, we develop a self-adaptive adversarial learning network which could further reduce the feature distribution shift through aligning the local feature distribution. Specifically, we design an adversarial network which is consisted of a feature extractor, a scenario discriminator, and two diverse classifiers. It could evaluate the degree of local feature distribution alignment by analyzing the prediction inconsistent of the classifiers. We design a self-adaptive adversarial loss which can be adjusted adaptively according to the degree of local alignment. If the features have been aligned locally, we reduce their impact on the loss to protect these aligned features. Otherwise, we increase their influence to accelerate the training process. The extensive experiments conducted on a designed mmWave testbed demonstrate that the proposed method could achieve an accuracy of at least 4% higher than those of existing cross-scenario DFGR methods, while the number of training iterations can be reduced by nearly half.

Power efficiency is a big concern in the Internet of Underwater Things (IoUT). The power consumption of underwater acoustic communications is typically in the scale of watts, which may drain the battery of underwater devices quickly. Whereas, the power consumption of underwater magnetic induction (MI) wireless communications is in the scale of milliwatt. Therefore, this article devotes to combine the underwater MI and acoustic communications to form a power-efficient underwater hybrid wireless network. Specifically, we investigate the power-efficient autonomous underwater vehicle (AUV) data collection schemes in an underwater MI and acoustic hybrid sensor network. We propose an alternating anchor nodes selection and flow routing (AANSFR) AUV data collection method, which alternately optimizes the AUV path planning and network data flow routing. The simulation results show that the proposed hybrid data collection scheme can significantly prolong the lifespan of underwater sensor networks.

Device-free gesture recognition is a potential noncontact human–computer interaction technique. It leverages the unique influence of the conducted gesture on surrounding wireless signals to accomplish gesture recognition. Existing methods usually leverage doppler spectrogram of the influenced wireless signals to characterize the motion pattern of gestures. These methods have achieved satisfactory accuracy when the gestures are conducted in a relatively fixed location, direction, and speed. However, when gestures are conducted in a different scenario, the recognition accuracy will drop dramatically. In this article, we try to solve this issue by characterizing the gesture motion pattern using a novel robust intrinsic spectrogram, which is independent of the conducted scenario. Specifically, we create a virtual coordinate system in which the coordinates of a gesture trajectory remain unchanged no matter where and how the gesture is conducted. Then, we design a coordinate transformation method to transform the raw doppler spectrogram into the robust intrinsic spectrogram to characterize the intrinsic motion pattern of the gesture. We further feed the intrinsic spectrogram into a deep network to realize gesture recognition. Extensive evaluations on a 77-GHz mmWave testbed show that the proposed method could achieve an average recognize accuracy of 88.4% with ten types of gestures.

Ship target detection in synthetic aperture radar (SAR) images is essential for many applications in marine monitoring and port security. Though considerable developments have been achieved, there still exist some issues toward multiscale and dense ship targets in complex inshore scenes. Under such common but challenging situations, it is difficult to extract effective target information, which drives the missing alarm rate rising dramatically. In complex scenes, it is hard to disentangle background noise from ship target information, which causes false alarm frequently. In this article, an anchor-free SAR ship detection method based on key-point estimation and attention mechanism is proposed to address the aforementioned issues. Specially, an anchor-free framework with skip connections and aggregation nodes is designed to fuse multiresolution features and detect multiscale ship targets. Moreover, a key-point estimation module is proposed to eliminate the undetected ship targets caused by dense target distribution. Furthermore, a channel attention module is explored to enhance network attention on ship targets and suppress background noise. Sufficient experimental results on the open SAR ship detection dataset demonstrate that compared with some state-of-the-art methods, the proposed method is able to achieve higher detection accuracy with a lower false alarm rate.

The visual loop-closure detection for autonomous underwater vehicles (AUVs) is a key component to reduce the drift error accumulated in simultaneous localization and mapping tasks. However, due to viewpoint changes, textureless images, and fast-moving objects, the loop closure detection in dramatically changing underwater environments remains a challenging problem to traditional geometric methods. Inspired by strong feature learning ability of deep neural networks, we propose an underwater loop-closure detection method based on a variational autoencoder network in this article. Our proposed method can learn effective image representations to deal with the challenges caused by dynamic underwater environments. Specifically, the proposed network is an unsupervised method, which avoids the difficulty and cost of labeling a great quantity of underwater data. Also included is a semantic object segmentation module, which is utilized to segment the underwater environments and assign weights to objects in order to alleviate the impact of fast-moving objects. Furthermore, an underwater image description scheme is used to enable efficient access to geometric and object-level semantic information, which helps to build a robust and real-time system in dramatically changing underwater scenarios. Finally, we test the proposed system under complex underwater environments and get a recall rate of 92.31% in the tested environments.

Multiaccess edge computing (MEC) empowers service providers (SPs) to run applications on the shared edge platforms in close proximity to mobile users, enabling ultralow latency access to a wide variety of cloud services. However, how to decide the amount of edge computing resources to rent for mobile service provisioning poses great challenges as the service demand is unknown to SPs a priori and may vary across the geographically distributed edge sites spatially and temporally. The resource rental decision also significantly affects SPs' deploying profits since it is critical for service deployment and workload assignment. This article investigates the service provisioning problem in a cooperative edge computing system under service demand uncertainty. We develop a holistic solution to make two-timescale decisions on edge resource rental and workload assignment to maximize SP's deploying profits. Briefly, we exploit historical service demand traces at the edge sites to characterize the uncertainty in a data-driven manner and formulate the edge service provisioning problem into a two-stage risk-averse optimization. To solve the formulated problem without compromising the data privacy, we propose an algorithm integrating Benders decomposition (BD) and alternating direction method of multipliers (ADMMs), which enables each edge site to keep the historical traces locally and participate in the optimization process. Based on real-world data sets, extensive simulations are conducted to validate the efficacy of our scheme.

Hyperspectral image classification, which is a crucial task for various remote sensing applications, can achieve qualified performance under a conventionalized assumption, i.e., there are sufficient samples in every concerned class. However, in field investigation, collecting enough samples for every class is extremely difficult, which results in defective training sets with insufficient or imbalanced samples and deteriorates the classification performance dramatically. In order to address this issue, we propose a hyperspectral image classification method based on a task-specific learning network. The proposed network works under an episode-based framework, which learns general knowledge from the tasks with sufficient samples by metalearning and relation learning and then inference specific knowledge for the tasks with few samples by parameters adjustment and representation comparison. In particular, a task-specific feature learner is designed to learn unbiased features, and a comparison-based classifier is utilized to adapt the minority classes. As a result, the proposed method can obtain a qualified overall accuracy over all samples with a comparative averaged accuracy over all classes. Extensive experiments on three real hyperspectral images show that our method can achieve state-of-the-art performance under both the few-shot and imbalanced training set settings.

The cross-data set knowledge is vital for hyperspectral image classification, which can reduce the dependence on the sample quantity by transferring knowledge from other data sets and improve the training efficiency by sharing knowledge between different data sets. However, due to the capturing environment change and imaging equipment difference, domain shift troubles the exploitation of the cross-data set knowledge. To address the aforementioned issue, this article proposes an unsupervised cross-data set hyperspectral image classification method based on adversarial domain adaptation. The proposed method, which employs multiple classifiers to build a discriminator and uses variational autoencoders to constitute a generator, works in an adversarial manner to drive the target samples under the support of the source domain. In particular, the classification error and the classification disagreement are considered in the objective function, which helps to align different domains while keeping the boundaries of different classes. Experimental results of the multidomain data set demonstrate that the proposed method can transfer and share cross-data set knowledge and achieve state-of-the-art performance without using the labeled information of the target data set.

The industrial Internet of Things (IIoTs) are well deployed to monitor pollutant emissions or device statuses in the industrial factory, especially in the chemical plants. To give a quick response of monitoring results, the industrial big data generated by IIoTs containing various tasks need to be processed as soon as possible. However, the priority constraints among the tasks generated by sensors are not considered in the existing architectures, which may result in the delayed response. Thus, in this article, we propose a mobile edge computing (MEC)-enabled architecture considering the priority constraints among tasks with the objective to minimize the response time. The tasks can be executed in MEC servers or cloud servers based on task complexity. Traditional methods search the optimal task allocation strategy through a set of initial strategies and a optimizer without the consideration of relationship among tasks. Thus, we propose a Bayesian network based evolutionary algorithm (BNEA) for optimizing a task allocation strategy. To fully consider the priority among tasks, the BNEA studies a Bayesian network based decomposition strategy in which the tasks are decomposed based on the relationship reflected by the learned Bayesian network structure. The BNEA searches the optimal task allocation strategy with the help of decomposed tasks cooperatively. Moreover, we propose a probability-based update strategy for particles to avoid draping into local optima. The experimental results verify that the BNEA can achieve the best response time through the corresponding task allocation strategy, which means that the data generated from the IIoTs can be transmitted to the destination in the shortest time.

According to the safety organization Kids and Cars, in US, an average of 38 children die each year in hot cars, seemingly forgotten by a distracted parent. Existing car seat alarm designs either compromise people's privacy (camera based designs), or fail to distinguish children sitting in the back from heavy stuff put on rear seats, and keep sending false alerts (pressure sensor based designs). In an effort to prevent such tragedies, we propose to utilize the fine-grained channel state information (CSI) from commercial off-the-shelf WiFi devices to detect if a child has been forgotten in rear seat of the car. Our child detection system only needs WiFi signal and applies both phase and amplitude measurement of the CSI. Based on this, our system can capture the movements of children, and effectively detect the children who are forgotten in rear seat and distinguish them from pets or other heavy stuff in rear seat with deep learning algorithms. In comparison with KNN based child detection method, the experiment results show that the performance of our deep learning based system increases dramatically, and the detection accuracy can reach more than 95%.

Nowadays, network intrusions have brought greater impact in a large scale. Intrusion Detection Systems (IDS) have been a recent research hotspot for both the industry and the academic. However, due to the dynamic characteristics of network traffic, it is challenging to extract significant features and identify the traffic types. This paper focuses on applying deep learning methods to feature extraction. Specifically, an IDS model is proposed based on autoencoder and long short-term memory (LSTM) cell. The overall architecture of the intrusion detection model includes a feature extractor, a classifier, and an evaluation block. Different structures of the feature extraction model have been discussed and researched. Experiments conducted on the UNSW-NB15 dataset produce satisfactory result. A number of selected metrics such as accuracy and false alarm rate are adopted to evaluate the detection performance. Simulation results indicate that our model works better than competing machine learning methods and achieves accuracy of over 92%.

Mobile crowdsensing (MCS) has become a new sensing and computing paradigm due to the proliferation of global positioning system (GPS) enabled mobile devices. There are three parties in the MCS, the MCS server, task requesters and workers. The MCS server needs to collect workers' location information to optimize the task allocation problem. However, during the location data collection process, workers' location privacy might be disclosed without their knowledge. It is challenging to preserve workers' location privacy while effectively and efficiently selecting proper workers to fulfill an MCS task. In this work, we propose a novel differentially private geocoding (DPG) mechanism to preserve workers' location privacy. Specifically, instead of reporting the exact latitude and longitude to the server, workers can use obfuscated geocode to describe their locations, since geocodes can provide an intuitive visualization of workers' spatial information to the MCS server. Based on the workers' obfuscated geocodes, we also formulate a travel distance minimization problem in MCS into an integer linear programming problem. We leverage conditional value at risk (CVaR) to characterize the uncertainty brought by the obfuscated geocodes, and develop feasible solutions to the formulated optimization problem. We conduct simulations with a real-world taxi dataset and verify the effectiveness of the proposed mechanism.

Ship detection from SAR images is an important remote sensing application. However, in complex scenes, i.e., the shore or harbor area, traditional ship detection methods cannot disentangle background information from the target ship, and the detection performance drops dramatically. Moreover, the severe coherent speckle noise also challenges ship detection from SAR images. In order to address the aforementioned issues, this paper proposes a SAR image ship detection method based on scene interpretation to improve the performance of ship detection in complex scenes. Firstly, segmentation algorithm based on Mask R-CNN is utilized to interpret the scene into two catalogs, i.e., the sea and the land. Then, ship detection algorithm based on Faster R-CNN is performed on the sea area and the land area respectively. Finally, non-maximum suppression is used to integrate detection results. Experimental results on SAR ship detection dataset illustrate that the proposed method produces high detection accuracy and low false alarm rate in complex scenes.

The explosive growth on emerging Internet-of-Things (IoT) applications makes our telecommunications networks confront twofold challenges. One is to provide sufficient flexibility for service diversity. The other is the shortage on spectrum. Network slicing and spectrum sharing have been deemed as two prominent solutions, which, however, are barely jointly studied in the literature. In this article, standing on both aspects, we propose a service-oriented spectrum-aware RAN-slicing trading (SSRT) scheme to achieve a dynamic on-demand RAN slicing under the spectrum sharing scenario. For the SSRT scheme, we jointly slice multidimensional resources, including heterogeneous spectrums (licensed and shared), time, and network facilities (nodes, radios, and powers). In particular, considering the uncertainty of shared spectrums, we distinguish them from the traditional licensed ones to fulfill different types of sessions, which are classified into delay tolerant and delay sensitive. To achieve an effective isolation, we construct a 4-D conflict graph and formulate the slice generation problem into a mixed-integer nonlinear programming (MINLP) problem, where a cross-layer resource allocation based on a hybrid transmission mode is designed for the customization. To cope with the difficulties when solving the problem, we employ the column generation algorithm to obtain the final slicing result over all the resources. The simulation results have shown the effectiveness of the proposed scheme.

Recent advances in device-free wireless sensing have created the emerging technique of device-free human gesture recognition (DFHGR), which could recognize human gestures by analyzing their shadowing effect on surrounding wireless signals. DFHGR has many potential applications in the fields of human-machine interaction, smart home, intelligent space, etc. State-of-the-art work has achieved satisfactory recognition accuracy when there are a sufficient number of training samples. However, it is time consuming and labor intensive to collect samples, thus how to realize DFHGR under a small training sample set becomes an urgent problem to solve. Motivated by the excellent ability of the generative adversarial network in synthesizing samples, in this article, we explore and exploit the idea of leveraging it to realize virtual samples augmentation. Specifically, we first design a single scenario network with new architecture and better-designed loss function to generate virtual samples using a few number of real samples. Then, we further develop a scenario transferring network to generate virtual samples by utilizing the real samples not only from the current scenario but also from another available scenario as well, which could improve the quality of synthesized samples with the extra knowledge learned from another scenario. We design an mmWave-based DFHGR testbed to test the proposed networks, extensive experimental results demonstrate that the augmented virtual samples are of high quality and facilitate DFHGR systems to achieve better accuracy.

Leveraging the mobility of autonomous underwater vehicles (AUVs) to collect and deliver data among different underwater devices enables numerous underwater Internet-of-Things (UW-IoT) applications. However, the most versatile underwater acoustic communications (UACs) may not be suitable in the AUV-assisted UW-IoT scenarios, considering the high cost and high power consumption of acoustic transducers, as well as high error rates of UACs due to the complex underwater acoustic channel conditions. Alternatively, we propose to apply the low-power magnetic induction (MI)-based wireless communications for AUV data dissemination and collection. Due to the mobility of AUVs and the underwater turbulence, MI channels between AUVs and other underwater devices are no longer stable and static, which poses great challenges to establish reliable MI links. To tackle this problem, we investigate the dynamic MI wireless communications in this article. We first mathematically characterize the dynamic MI channel when an AUV approaches its target for data collection. Based on this dynamic channel model, the dynamic communication range and available bandwidth of MI are derived. We also build an MI wireless communication system that can work within a dynamic range. The communication performances are evaluated through numerical simulations as well as underwater experiments.

Device-free gesture recognition (DFGR) is a promising sensing technique, which can recognize a gesture by analyzing its influence on surrounding wireless signals. Most of the DFGR systems are designed based on machine learning. However, the recognition performance will drop dramatically when the testing condition is different with the training one. Inspired by the transferrable knowledge learning ability of humans, this paper develops a practical DFGR system based on metalearning to solve the aforementioned problem. Specifically, we design a deep network which could not only learn discriminative deep features, but also learn a transferrable similarity evaluation ability from the training set and apply the learned knowledge to the new testing conditions. Extensive experiments conducted by four users in two scenarios demonstrate that the proposed system could recognize new types of gestures, or gestures performed in new conditions, with an accuracy of more than 90%, using very few number of new samples.

This paper proposes a peer-assisted power supply approach for a wireless powered Internet of Things (IoT) non-orthogonal multiple access (NOMA) network in which passive user equipments (UEs) without battery harvest energy from active UEs equipped with power supply. Specifically, passive UEs harvest energy from active UEs during their uplink transmission using NOMA. They then upload information along with the active UEs. Particularly, considering the combination of time division multiple access (TDMA) and NOMA, under the assumption that the power of active UEs is fixed, we study different transmission modes (non-stand-alone/stand-alone) and different operations (NOMA/NOMA-plus-TDMA). Taking into account the practical applications, we re-investigate the above schemes in the scenario where active UEs' energy is limited, i.e., the power of active UEs is not fixed and is affected by time allocation. We maximize the sum-throughput of each proposed model. We prove that the optimization problems for all cases are convex, and we obtain closed-form solutions for most cases. Finally, we show by simulations that, in all cases, the transmit power of active UEs and the number of UEs have a positive effect on the sum-throughput. Besides, in terms of maximizing the sum-throughput, the NOMA-plus-TDMA operation outperforms the NOMA operation. If active UEs' power is fixed, the non-stand-alone transmission outperforms the stand-alone transmission, and vice versa.

Network Intrusion Detection Systems (NIDS) have become a strong tool to alarm attacks in computer and communication systems. Machine learning, especially deep learning, has made huge success in fields of industry and academic. Network intrusion activity can be a time series event. In this paper, we adopt a time-related deep learning approach to detect network intrusions. A stacked sparse autoencoder (SSAE) is first built to extract the features with the greedy layer-wise strategy. And then, we propose a time- related intrusion detection system based on the variants of Recurrent Neural Network (RNN). We study the performance of proposed approach on the binary classification with a benchmark dataset UNSW- NB15. Based on the study of parameter time steps, it is proved that our time- related model is effective for intrusion detection. The experiment results show that the accuracy of the proposed approach reaches over 98% and the false alarm rate is as low as 1.8%. The performance of our model is superior to that of the standard RNN- based approach and approaches based on Deep Neural Network and shallow machine learning.

Due to the rapid development of learning-based methods, hyperspectral image classification has achieved remarkable progress. However, since the semantic discrepancy between different land-cover types, the feature distributions of different classes are so nonuniform that the classifier can not measure them with a single rule. In order to consider semantic knowlage in classification, this paper propose a knowledge guided classification method based on hierarchical class tree and deep learning. The proposed method fuses similar classes into several super classes by the knowledge of the confusion matrix, and classify multi-level super classes with different deep networks. Extensive experiments on two hyperspectral images demonstrate that the proposed method can utilize semantic information of different land-cover types, and give better performance than traditional methods.

With the rapid growth on data traffic, spectrum shortage becomes increasingly serious, leading to the paradigm shift in spectrum usage from an exclusive mode to a sharing mode. However, how to utilize shared spectrums effectively for service provisioning is not straightforward due to its uncertain availability, known as spectrum uncertainty. In this paper, we propose a new metric to evaluate the achievable rate of a link on a share band under a confidence level, called probabilistic link capacity, which offers us an effective way to guarantee the quality of service statistically when using the shared spectrum for service delivery. Different from most existing works where the distributional information is explicitly given based on certain structural assumption, we develop a data-driven distributionally robust approach by using the first and second order statistical information. To achieve the result, we formulate it into a tractable semidefinite programming problem based on the worst-case of conditional-value-at-risk. Finally, as a use case, we design a service-based spectrum-aware transmission scheme, so that different kinds of spectrums (licensed and shared) can be efficiently utilized to satisfy the diverse service requirements.

One of the key problems in spectrum sensing is to design the test statistic. Existing methods generally exploit the model-based features as the test statistic, such as energies and eigenvalues. However, these features could not accurately characterize the real environment. Motivated by this, in this paper, we use a deep neural network (DNN) to intelligently explore the data-driven test statistic. Firstly, we introduce a DNN-based detection framework, where a DNN-based likelihood ratio test (DNN-LRT) is derived to guarantee the optimality of the designed test statistic. As a realization of the developed DNN-based framework, we use the sample covariance matrix as the input of a convolutional neural network (CNN), and propose a covariance matrix-aware CNN (CM-CNN)-based spectrum sensing algorithm, which further improves the performance. In addition, we also provide the theoretical analysis of the proposed method. To the best of our knowledge, it's the first time to analyze the theoretical performance of CNN-based methods. Finally, simulation results demonstrate that the performance of the proposed method is close to that of the optimal detector. Particularly, the proposed method could achieve a detection probability of 96.7% with a false alarm probability of 1.9% at SNR = -18dB, which significantly outperforms the conventional methods.

Deep learning-based classification methods are competent to achieve an excellent performance under one necessary condition, i.e., there are sufficient labeled samples in each class, which is extremely impractical in most of the remote sensing tasks. To improve the performance with small training sets, we resort to other hyperspectral images and design a two-phase relation learning network that can be transferred between different images for general information sharing and fine-trained on a specific hyperspectral image for individual information learning. Specifically, we use a relation learning method to compare samples and deal with the task inconsistency between different data sets, and we adopt an episode-based training strategy to mimic the testing setup and learn the transferable comparison ability. Benefited from these two strategies, the proposed network takes the advantage of extra knowledge for information supplement and learns to compare rather than to classify for information exploration, which guarantees a reasonable performance even with small training sets. Extensive experiments and analysis on three benchmarks demonstrate that the proposed method can provide an effective solution for hyperspectral image classification with small training sets, which makes it possible to work on large-scale applications of earth observation with less effort on field investigation.

For hyperspectral image classification, there is a large gap between the theoretical method and the practical application. Hyperspectral image classification in theoretical research trains a new classifier for each data set, which is ineffective and even infeasible in large-scale applications. In this paper, we make a preliminary attempt to recycle the classification model to new data sets in an unsupervised way. Specially, we propose a cross-data set hyperspectral image classification method based on deep domain adaptation. The proposed method contains three modules: domain alignment module that learns to minimize the domain discrepancy with the guide of an irrelevant task, task allocation module that learns to classify on the source domain with the regulation of domain alignment, and domain adaptation module that transfers both the alignment ability and classification ability to the target domain by an adaptation strategy. As a result, with the information of an irrelevant task on dual-domain data sets, we can minimize the domain discrepancy and transfer the task-relevant knowledge from the source domain to the target domain in an unsupervised way. Extensive experiments on three hyperspectral images demonstrate the effectiveness of our method compared with other related methods when dealing with new data sets.

Device-free activity recognition (DFAR) is a promising technique that detects the activity of a target by analyzing the influence of its existence on surrounding wireless links. It realizes target sensing without the participation or even awareness of the target. The key question of DFAR is how to characterize the influence of the target on wireless links. Existing works mostly utilize statistical features, such as mean and variance in time-domain, and energy as well as entropy in frequency-domain, to characterize the influenced signals. However, statistical features provide only partial information. This paper explores the method on how to characterize the distribution of the signal as a whole. Specifically, we present a novel coherence histogram, which leverages the spatial structural characteristics to better characterize the distribution of the wireless signal. The coherence histogram captures not only the occurrence probability of received signal strength (RSS) measurements, but also the spatial relationship between adjacent RSS measurements as well. Experimental results show that our coherence histogram-based DFAR system could achieve an accuracy of more than 96%, which significantly outperforms other state-of-the-art DFAR systems remarkably.

This paper designs a novel geometry-conformal antenna for Magnetic Induction (MI)-based subsea wireless communications for autonomous underwater vehicles (AUV). The designed tri-directional antennas can be wrapped directly on the surface of AUVs, such that the AUVs fluid dynamics are well maintained to ensure power efficiency of the vehicles. In addition, ferrite materials are added between the MI antenna and the metallic body surface of the AUVs to overcome the shielding effect and enhance the MI signal strength. The designed MI communication system is implemented in hardware and the effectiveness of the geometry-conformal MI antenna is demonstrated through COMSOL simulations and lab experiments.

Due to its promising application in the field of human- machine interaction, letter recognition in the air has drawn considerable attention in recent years. Compared with traditional sensor-based and camera-based methods, letter recognition in the air using channel state information (CSI) is more user-friendly and easy-to-deploy. Unfortunately, due to the limited range of the moving hand and the similarity of different letters, it is difficult to extract discriminative writing patterns for different letters from the noisy environment. In this paper, we design LetFi, a high accuracy letter recognition in the air system, which could detect and recognize the letter written by a user by analyzing its influence on surrounding WiFi signals. Specifically, we design a differential method to extract robust CSI measurements, develop a variance based scheme to detect the start and the end points of the letter writing activity, and propose a coherence histogram based multi-domain feature extraction strategy to extract discriminative features from not only the time domain and frequency domain, but also the spatial structural domain. Extensive experimental results show that the proposed LetFi system could achieve a recognition accuracy of 95% when recognizing the 26 capital letters.

Radio-Frequency Identification (RFID) nowadays has been widely used in a variety of fields such as manufacturing, transportation and healthcare, for identification, tracking, and information collection due to the simplicity and low cost of RFID tags. Existing RFID information collection protocols focus on information collection at a single reader, which aim to minimize the collision among RFID tags and the execution time for information collection. However, none of them consider reader scheduling to deal with the collision between neighboring readers that exists in multi-reader RFID systems. In this paper, we propose two reader scheduling schemes for time-efficient information collection in multi-reader RFID systems. The two proposed scheduling schemes can coordinate multiple readers to collect information simultaneously and ensure information from all tags is successfully collected with reader-reader collision. Extensive simulation results demonstrate the efficiency of the proposed reader scheduling schemes for multi-reader information collection.

Convolution neural network (CNN) utilizes alternating convolutional and pooling layers to learn representative spatial information when the training samples are sufficient. However, for pixelwise classification of hyperspectral image, some important information is neglected by CNN, such as the erased information by the pooling operation and the appearance information from lower layers. Moreover, the lack of training samples is a common situation in remote sensing area, which afflicts CNN with overfitting problem. To address the aforementioned issues, this paper designs an end-to-end deconvolution network with skip architecture to learn the spectral-spatial features. The proposed network starts with two branches, i.e., the spatial branch and spectral branch. In the spatial branch, a band selection layer is designed to reduce parameters and remit the overfitting problem, unpooling and deconvolution operations are utilized to recover the erased information of the pooling layers and learn pixelwise spatial representation hierarchically, and the skip architecture is constructed for merging the deep semantic information with the shallow appearance information. In the spectral branch, a contextual deep network is employed for learning deep spectral features. Experimental results on three benchmark data sets reveal the competitive performance of the proposed approach over several related methods.

Spectrum trading creates more accessing opportunities for secondary users (SUs), and economically benefits the primary users (PUs). Compared with centralized spectrum trading designs, e.g., spectrum auction, distributed spectrum trading captures instantaneous spectrum trading opportunities better over large geographical regions without incurring extra infrastructure deployment and has no network scalability issues. However, the existing distributed spectrum trading designs have limited concern regarding spectrum reuse. Considering spatial reuse, in this paper, we propose a novel distributed frequency reuse-based opportunistic spectrum trading (D-FROST) scheme, which can further improve spectrum utilization, provide more accessing opportunities for SUs, and increase the revenues of PUs. In this paper, we employ conflict graph to characterize the SUs' co-channel and radio interferences, and mathematically formulate a centralized PUs' revenue maximization problem under multiple wireless transmission constraints. Due to the NP-hardness to solve the problem and the non-existence of centralized trading entity, we develop the D-FROST algorithms based on matching with evolving preferences, and prove its stability. Through extensive simulations, we show that the proposed D-FROST algorithm is superior to other distributed spectrum trading algorithms without considering spectrum reuse, yields results close to the centralized optimal one, and is effective in increasing PUs' revenue and improving spectrum utilization.

Recent advances in device-free wireless sensing (DFS) have shown that it may eventually evolve traditional wireless networks into smart networks which could sense surrounding target location and activity information without equipping the target with any devices. Despite its promising application prospects, one challenging problem to be solved is that the performance of the DFS system degrades significantly in complex scenarios, such as through-wall and non-line-of-sight (NLOS) scenarios. To alleviate this problem, this paper seeks to explore and exploit more informative features from not only the time domain and frequency domain, but also the spatial structural domain. We partition the time domain and frequency domain measurement matrices into basic structure blocks, adopt self-organizing map networks to cluster the blocks into a number of categories, so as to make it feasible to characterize the block distributions. We further adopt coherence histograms to characterize the distribution of the blocks by considering the spatial relationship between adjacent blocks. Thanks to the additional information provided by the spatial structural domain, extensive experimental results achieved in through-wall and NLOS scenarios confirm the outstanding performance of the proposed multi-domain features based DFS system.

State-of-the-art sensing methods mostly exploit spectrum holes (SHs) in conventional frequency, time, and geography dimensions, which can hardly satisfy the increasing throughput demand of CR networks. Meanwhile, the rapid development of multi-antenna technology makes the terminal obtain the angle recognition capability. Motivated by this, this paper analyzes SHs from the angle/space domain and design a spatial sector based sensing-access scheme. In this case, the SH can be regarded as a kind of particle in spatial-temporal dimension and thus the spatial-temporal uncertainty principle (STUP) is discovered, which reveals an interesting constraint phenomenon between spatial and temporal resolutions. Based on STUP, we propose a novel spatial-temporal resolution tradeoff (STRT) scheme, whose objective is to identify the optimal spatial resolution size to maximize the throughput of CR networks. Different from the conventional temporal domain sensing-throughput tradeoff problem, we study the spatial-temporal cross-dimension optimization, thus fully exploiting SHs' spatial diversity to achieve a better performance. In addition, a fast search algorithm is proposed to track the optimal spatial resolution at an exponential convergence rate. Simulation results verify the efficiency of the proposed tradeoff scheme and search algorithm.

Cognitive radio networks (CRNs) have been emerging as a promising technique to improve the spectrum efficiency of wireless and mobile networks, which form spectrum clouds to provide services for unlicensed users. As spectrum clouds, the performance of multi-hop CRNs heavily depends on the routing protocol. In this paper, taking the newly proposed Cognitive Capacity Harvesting network as an example, we study the routing problem in multi-hop CRNs and propose a spectrum-aware anypath routing (SAAR) scheme with consideration of both the salient spectrum uncertainty feature of CRNs and the unreliable transmission characteristics of wireless medium. A new cognitive anypath routing metric is designed based on channel and link statistics to accurately estimate and evaluate the quality of an anypath under uncertain spectrum availability. A polynomial-time routing algorithm is also developed to find the best channel and the associated optimal forwarding set and compute the least cost anypath. Extensive simulations show that the proposed protocol SAAR significantly increases packet delivery ratio and reduces end-to-end delay with low communication and computation overhead, which makes it suitable and scalable to be used in multi-hop CRNs.

The state-of-the-art eigenvalue-based spectrum sensing methods only consider the partial information of eigenvalues, such as the maximum, minimum, and mean values to make detection, which does not make full use of the eigenvalues to catch correlation. In this paper, we focus on all the eigenvalues of sample covariance matrix in multi-antenna cognitive radio networks and propose eigenvalue weighting-based detection schemes. According to the Neyman–Pearson criterion, the globally optimal weighting solution is the likelihood ratio test (LRT). Hence, we analyze and derive the eigenvalue-based LRT (E-LRT). Utilizing the random matrix theory, a simple closed-form expression for the E-LRT is obtained, which is exactly the optimal eigenvalue weighting scheme. Although the E-LRT is optimal, it is infeasible in practice due to its dependence on the knowledge of primary users and noise powers. Hence, we further analyze suboptimal methods and design maximum likelihood estimation-based approximation weighting approach. Under the approach, both semi-blind (only the noise power is known) and totally-blind methods are correspondingly proposed. In addition, the theoretical performance analysis of these proposed methods are provided. Simulation results are presented to verify the efficiency of the proposed algorithms.

Context awareness plays an important role in many emerging applications, such as mobile computing and smart space. Since FM signal is ubiquitous, it has been recognized as an attractive and promising technique to realize context awareness. When a target is at different locations or performs different activities, it will exert different influence on the FM signal around it. Therefore, it is possible to deduce its location and activity by analysing its influence on the FM signal. However, FM signal is extremely weak and noisy, which makes it a challenging task to achieve high-performance context awareness. In this paper, we propose a new method for improving the performance of an FM-based context-aware system using multidomain features. Specifically, we extract signal features not only from the time domain, but also from the wavelet domain, the frequency domain, and the space domain, and construct robust and discriminative multi-domain features to characterize the FM signal. Furthermore, we also model context awareness as a classification problem and develop a robust iterative sparse representation classification algorithm to efficiently solve this problem. Extensive experiments performed in a 7.2m×10.8m clutter indoor laboratory with one multi-channel FM receiver demonstrate that the proposed schemes could achieve more than 90% accuracy of location estimation and activity recognition when 3 antennas are used.

This paper presents a hyperspectral image classification method based on deep network, which has shown great potential in various machine learning tasks. Since the quantity of training samples is the primary restriction of the performance of classification methods, we impose a new prior on the deep network to deal with the instability of parameter estimation under this circumstances. On the one hand, the proposed method adjusts parameters of the whole network to minimize the classification error as all supervised deep learning algorithm, on the other hand, unlike others, it also minimize the discrepancy within each class and maximize the difference between different classes. The experimental results showed that the proposed method is able to achieve great performance under small training set.

Device-free wireless localization (DFL) is a promising technique. It can localize and track a target without carrying any electronic device. Compared with traditional received signal strength (RSS)-based DFL, the recently proposed time-of-flight (TOF)-based DFL technique could achieve better performance. However, TOF-based DFL requires that each pair of nodes should perform a pairwise TOF measurement sequentially, which makes it impractical for being utilized in time and energy sensitive applications. Inspired by the fact that a target shadows only a small subset of wireless links, which travel through its spatial impact area, we incorporate link state estimation function into particle filter (PF) framework to predict shadowed links and design a parallel scanning scheme to scan multiple shadowed links simultaneously. The aforementioned methods guarantee that the effective TOF measurements can be acquired timely and energy efficiently. Meanwhile, to achieve reasonable localization performance with these reduced-under-sampled link measurements, we utilize compressive sensing (CS) algorithm to reconstruct the shadowing effect map (SEM). With the map as observation likelihood function, PF algorithm could estimate targets location accurately. Experimental results with an 802.15.4a chirp spread spectrum ranging hardware testbed show that the average running time and energy consumption reduce remarkably.

The robustness of inter‐session network coding is still an open issue in lossy wireless networks. The traditional XOR based network coding cannot work well if the overhearing is unperfect. Especially, the coding node cannot know the overheard information in time. In this paper, we consider a robust network coding method, namely R 2 NC which uses random linear network coding to encode packets together in the inter‐session level, to resist the unperfect overhearing problem. With this method, coding node can always know the solvability of coded packets without the knowledge of overheard information. We analyse the performance of R 2 NC method with both lossy links of output and overhearing in the classic X ‐topology model, and give a necessary condition for the existence of coding gain. Finally, we design an optimal coding algorithm and a relay selection algorithm for R 2 NC to achieve its maximal transmission efficiency. Through ns‐2 simulations, we demonstrate that R 2 NC plays a good performance in terms of throughput, delay and overhead, and is robust against losses on output and overhearing links.

For realizing robust target tracking with wireless sensor networks in the circumstance where the propagation parameters of the characteristic signal emitted by the target are unknown, a novel tracking algorithm under the particle filter framework is proposed. We propose a scheme to realize particle weight calculation without the prior knowledge about the propagation parameters of the target's characteristic signal. With the use of the monotonic relationship of the distance and the received signal strength, we define the signal characteristic sequence and particle distance sequence and utilize the modified sequence distance between the signal characteristic sequence and the particle distance sequence as the criterion to calculate the particle weight blindly with simple lightweight operations. Simulation results demonstrate the effectiveness of the proposed algorithm. Copyright © 2011 John Wiley & Sons, Ltd.

Due to its ability of realizing device free localization with wireless networks, the radio tomography becomes a promising technique that draws considerable attention. Traditional radio tomography makes use of the received signal strength (RSS) of wireless links to realize location estimation. However, the RSS measurement is particularly sensitive to noise. Inspired by the fact that similar to the RSS, the time-of-flight (TOF) measurement also changes significantly when some objects shadow the wireless link, and the fact that compared with the RSS, the TOF measurement is robust to noise, a novel TOF-based radio tomography is proposed in this paper. With the TOF measurements of the shadowed links as observation information, a modified particle filter algorithm which utilizes the compressive sensing technique to produce the importance distribution of the particle set is proposed, so as to realize localization and tracking with under-sampled measurements by making full use of the space-domain sparse and time-domain gradually changed feature of the location information. The experiments with the 802.15.4a chirp spread spectrum ranging hardware are presented to confirm the proposed scheme.

A compressive sensing-based approach to solve the problem of tracking targets in the deployment area of the wireless networks without the need of equipping the target with a wireless device has been proposed. We present a dynamic statistical model for relating the change of the received signal strength between the node pairs to the spatial location of the target. On the basis of the model, the problem is formulated as a sparse signal reconstruction problem, and we propose a novel Bayesian greedy matching pursuit (BGMP) algorithm to tackle the signal reconstruction problem even from a small set of measurements. The BGMP iteratively seeks the contribution of each pixel for multi-times to compensate for the inaccuracy of the measurement matrix, and builds the enumeration region based on the past estimations to speed up the algorithm and improve its reconstruction performance simultaneously. Experimental results demonstrate the effectiveness of our approach and confirm that the BGMP algorithm could achieve satisfactory localisation and tracking results.

Abstract Benefitting from its ability to estimate the target state's posterior probability density function (PDF) in complex nonlinear and non‐Gaussian circumstance, particle filter (PF) is widely used to solve the target tracking problem in wireless sensor networks. However, the traditional PF algorithm based on sequential importance sampling with re‐sampling will degenerate if the latest observation appear in the tail of the prior PDF or if the observation likelihood is too peaked in comparison with the prior. In this paper, we propose an improved particle filter which makes full use of the latest observation in constructing the proposal distribution. Thequality prediction functionis proposed to measure the quality of the particles, and only the high quality particles are selected and used to generate the coarse proposal distribution. Then, acentroid shift vectoris calculated based on the coarse proposal distribution, which leads the particles move towards the optimal proposal distribution. Simulation results demonstrate the robustness of the proposed algorithm under the challenging background conditions. Copyright © 2010 John Wiley & Sons, Ltd.