Spiking Neural Networks (SNNs) transmit information via discrete spike events, offering advantages in energy efficiency and computational cost. However, current SNN-based object detectors suffer from limited feature expression and inefficient fusion due to temporal sparsity and the asynchronous nature of spike features. To address these challenges, we propose SpikeGate-YOLO, a spiking object detection architecture optimized for spike-driven processing. Specifically, we introduce the Reparam-Spike Gating (RSG) block to enhance feature expressiveness while maintaining computational efficiency. We also design the Spike Multi-Granularity Difference-aware Feature Harmonizer (SpikeMDFH), which improves multi-scale feature fusion through dynamic attention and biologically inspired gating mechanisms, preserving spike sparsity. Experiments on both the COCO and Gen1 datasets show that SpikeGate-YOLO achieves state-of-the-art results, reaching 63.4% mAP@50 and 46.3% mAP@50:95 on COCO, and 69.3% mAP@50 and 42.9% mAP@50:95 on Gen1. These results confirm the effectiveness of our architecture in overcoming spike-specific limitations in feature representation and fusion for object detection.

In mining engineering, the mine hoist plays a vital role in transporting materials between the surface and under-ground. The wire rope, serving as the main connecting element, is particularly critical. While current safety monitoring often utilizes computer vision, underground environments pose chal-lenges like poor lighting, which can compromise camera accuracy. To address this issue, this paper introduces a millimeter-wave monitoring approach. This method achieves precise measurement of the wire rope's swing distance and speed by accurately extracting swing signals from a single millimeter-wave channel and optimizing radar parameters. Additionally, a reliable multi-view fusion model is developed to integrate decision-making on swing distance and speed characteristics. This approach improves the reliability and accuracy of hoist monitoring, even in complex mining environments.

As cybersecurity risks increase, reliable user authentication has become crucial. Traditional static methods, such as facial recognition, are vulnerable to data hijacking threats. This paper presents a novel new paradigm for CAPTCHA (Completely automated public turing test to tell computers and humans apart) verification, dynamic gesture, aimed at enhancing security and robustness. By integrating visual and inaudible sound signals across two complementary dimensions, this approach reduces blind spots and increases the cost of spoofing for CAPTCHA verification. Additionally, a trustworthiness fusion network is introduced, which incorporates a modality trustworthiness calculation method based on Dirichlet distribution, and factors of information entropy and distance depth, enabling dynamic decision-making, significantly improving accuracy and adaptability. Experimental results demonstrate the method’s practical feasibility and achieve an accuracy of 97% in distinguishing between humans and bots.

Achieving high-precision, universal localization services remains a significant challenge, as existing solutions typically rely on specialized hardware or complex algorithms. This paper aims to develop a lightweight and ubiquitous localization scheme that utilizes commercial audio devices (two speakers and a microphone). We control the two speakers to transmit the Orthogonal Frequency Division Multiplexing (OFDM) signals within the same frequency band, creating a composite interference fields formed by multiple subcarriers. Our main observation is that the initial phase difference between coherent signals leads to a spatial shift of the interference sound field. Therefore, we design a phase modulation mechanism that applies unique initial phase differences to each pair of subcarriers, producing an asymmetric interference sound field that provides an interference intensity distribution with significant spatial diversity. Finally, based on the intensity information recorded by the microphone, we construct the Multi-subcarrier Interference Intensity (MII) curve and propose effective curve matching method for location estimation. Extensive simulations and experiments have verified the effectiveness of the proposed method, and the median localization accuracy in real environments can reach 3.21 cm.

Softassign is a pivotal method in graph matching and other learning tasks. Many softassign-based algorithms ex-hibit performance sensitivity to a parameter in the softas-sign. However, tuning the parameter is challenging and al-most done empirically. This paper proposes an adaptive softassign method for graph matching by analyzing the re-lationship between the objective score and the parameter. This method can automatically tune the parameter based on a given error bound to guarantee accuracy. The Hadamard-Equipped Sinkhorn formulas introduced in this study signif-icantly enhance the efficiency and stability of the adaptive softassign. Moreover, these formulas can also be used in optimal transport problems. The resulting adaptive softas-sign graph matching algorithm enjoys significantly higher accuracy than previous state-of-the-art large graph matching algorithms while maintaining comparable efficiency.

We propose a Dynamic Interaction Graph Convolutional Network (DIGCN), an image object detection method based on learnable proposals and GCN. Existing object detection methods usually work on dense candidates, resulting in redundant and near-duplicate results. Meanwhile, non-maximum suppression post-processing operations are required to eliminate negative effects, which increases the computational complexity. Although the existing sparse detector avoids cumbersome post-processing operations, it ignores the potential relationship between objects and proposals, which hinders detection accuracy improvement. Therefore, we propose a dynamic interaction GCN module in the DIGCN, which performs dynamic interaction and relational modeling on the proposal boxes and proposal features to improve the object detection accuracy. In addition, we introduce a learnable proposal method with a sparse set of learned object proposals to eliminate a huge number of hand-designed object candidates, avoiding complicated tasks such as object candidate design and many-to-one label assignment, and reducing object detection model complexity to a certain extent. DIGCN demonstrates accuracy and run-time performance on par with the well-established and highly optimized detector baselines on the challenging COCO dataset, e.g. with the ResNet-101FPN as the backbone our method attains the accuracy of 46.5 AP while processing 13 frames per second. Our work provides a new method for object detection research.

Seismograms, the fundamental seismic records, have revolutionized earthquake research and monitoring. Recent advancements in deep learning have further enhanced seismic signal processing, leading to even more precise and effective earthquake monitoring capabilities. This paper introduces a foundational deep learning model, the Seismogram Transformer (SeisT), designed for a variety of earthquake monitoring tasks. SeisT combines multiple modules tailored to different tasks and exhibits impressive out-of-distribution generalization performance, outperforming or matching state-of-the-art models in tasks like earthquake detection, seismic phase picking, first-motion polarity classification, magnitude estimation, back-azimuth estimation, and epicentral distance estimation. The performance scores on the tasks are 0.96, 0.96, 0.68, 0.95, 0.86, 0.55, and 0.81, respectively. The most significant improvements, in comparison to existing models, are observed in phase-P picking, phase-S picking, and magnitude estimation, with gains of 1.7%, 9.5%, and 8.0%, respectively. Our study, through rigorous experiments and evaluations, suggests that SeisT has the potential to contribute to the advancement of seismic signal processing and earthquake research.

Graph Neural Networks (GNNs) have achieved great success in graph representation learning. Modern GNNs are built upon the homophily assumption and iteratively aggregate messages from immediate neighbors through the message-passing mechanism, which limits the ability of GNNs to represent graphs with heterophily. Existing GNNs considering heterophily adapt to the heterophilous graphs through reconstructing neighborhood and message fusion. However, noise information is constantly superimposed during message passing, due to the mixed propagation of different-order messages. In this paper, we propose a special graph convolutional network with virtual edge message passing (VEGCN), which consists of three important components: virtual edge message passing, message attention, and residual connection. Different from the message-passing mechanism of existing GNNs, virtual edge message passing can directly transfer messages from the second-order neighbors to the target nodes. By bypassing the first-order neighbors, VEGCN can avoid interference from first-order neighbors. In addition, we design an attention mechanism to adaptively obtain messages from first-order and second-order neighbors. This attention mechanism can distinguish the different importance of the messages from first-order neighbors and second-order neighbors. Finally, we introduce a residual connection to enhance the features of the nodes themselves and alleviate over-smoothing. We validate the effectiveness of VEGCN on several benchmark datasets including graphs with homophily and heterophily. Experimental results show that VEGCN outperforms representative baselines. Furthermore, we also designed ablation experiments to verify the role of the core components.

Early symptom monitoring is an essential measure for pneumoconiosis prevention. However, one severe limitation is the high requirement for a dedicated device. This paper proposes$p^{3}Warning$to realize low-cost warnings for potential pneumoconiosis patients via contactless sensing. For the first time, the designed framework utilizes the inaudible acoustic signal with a pair of commercial speaker and microphone to monitor early symptoms of pneumoconiosis including abnormal respiration and cough. We introduce and address unique technical challenges, such as designing a delay elimination method to synchronize transceiver signals and providing a search-based signal variation amplification strategy to support highly accurate and long-distance vital sign sensing. Comprehensive experiments are conducted to evaluate$p^{3}Warning$. The results show that it can achieve a median error of 0.52 bpm for abnormal respiration pattern monitoring and an accuracy of 95 % for cough detection in total, and support the furthest range of up to 4 m.

Fall detection is an essential measure for the safety of elders. While traditional contact-based methods support acceptable detection performance, the recent advance in wireless sensing could enable contact-free fall detection. However, two severe limitations are short sensing range and weak through-wall capability, which hampers wide applications in smart homes. This paper proposes a novel system LoFall, which is the first time to utilize the LoRa signal to realize contact-free long-range through-wall fall detection. We address unique technical challenges, such as proposing a novel strategy of candidate signal search to reduce the calculation time of fall detection and designing a weighted feature fusion algorithm based on fuzzy entropy to improve the accuracy of through-wall fall detection. Comprehensive experiments are conducted to evaluate LoFall. Results show that it can achieve a total accuracy of 93.3% for through-wall fall detection, and support the furthest detection range of up to 10 m.

Hand gesture recognition is an essential Human Computer Interaction (HCI) mechanism for users to control smart devices. While traditional device-based methods support acceptable recognition performance, the recent advance in wireless sensing could enable device-free hand gesture recognition. However, two severe limitations are serious environmental interference and high-cost hardware, which hamper the wide deployment. This paper proposes a novel system TaGesture, which employ the inaudible acoustic signal to realize device-free and training-free hand gesture recognition with a pair of commercial speaker and microphone array. We address unique technical challenges, such as proposing a novel acoustic hand tracking smoothing algorithm with Interaction Multiple Model (IMM) Kalman Filter to address the issue of localization angle ambiguity, and designing a classification algorithm to realize acoustic-based hand gesture recognition without training. Comprehensive experiments are conducted to evaluate TaGesture. Results show that it can achieve a total accuracy of 97.5% for acoustic-based hand gesture recognition, and support the furthest sensing range of up to 3 m.

Respiration rate is an essential vital indicator for health monitoring. While traditional sensor-based methods support acceptable sensing performance, the recent advance in wireless sensing could enable sensor-free and contact-free respiration sensing, which is particularly important during the practice of social distancing against a pandemic like COVID-19. Among a variety of wireless technologies employed for respiration sensing, Wi-Fi-based solutions are most popular due to the pervasive development of infrastructure. However, the existing Wi-Fi-based approaches need to retrieve Wi-Fi readings from access points, which are not often accessible for the end users. In this article, we propose a novel system, MoBreath, in which we utilize the Wi-Fi channel state information (CSI) readings extracted from the end-user device, a smartphone, to monitor the respiration rate for the first time. We introduce and address unique technical challenges, such as selecting the optimum CSI subcarriers from many noisy candidates and providing smartphone placement strategies for both single and multiple human target scenarios based on the Fresnel zone model to support highly accurate respiration sensing. Our evaluation of MoBreath using commodity smartphones in different environments shows that it can accurately estimate the respiration rate at a low error rate of 0.34 breaths per minute and support the sensing range of up to 3–4 m. Even for challenging scenarios such as the target is covered by a quilt and multiple targets are in the sensing area, MoBreath can still support highly accurate results.

Aiming at problems of under-fitting and poor model robustness in learning-based wireless sensing methods caused by the lack of large-scale wireless sensing datasets, this paper proposes a privacy-friendly collaborative wireless sensing framework, called Co-Sense. It builds a community with multiple clients and a server, which aggregates the clients' local models into a federated model with cross-domain capability. To protect the privacy of users' local data, we innovatively introduce the idea of federated learning into the field of wireless sensing, by uploading users' local model parameters instead of their local data. Then, in response to the uneven computing power of different users' edge devices, we propose a local model update algorithm based on adaptive computing power. Furthermore, a client selection algorithm based on test nodes is designed to reduce the negative influence of malicious clients on Co-Sense. Finally, we evaluate Co-Sense on three well-known public wireless datasets, including the gesture dataset, the activity dataset, and the gait dataset. Experimental results show that the sensing accuracy of Co-Sense is more than 10% higher than that of the most advanced wireless sensing models.

COVID-19 outbreaks rapidly around the world, which is the enemy faced by all humankind. Since COVID-19 is mainly spread through close personal contact, searching close-contacts is key to controlling this virus's spread. This paper designs COVID-19 Tracer, a novel low-cost passive system for searching COVID-19 patients' close-contacts. Utilizing ubiquitous Wi-Fi probe requests, COVID-19 Tracer can quickly determine whether a person stays in one small space with a COVID-19 patient in the same period. Furthermore, it seeks to find out a close-contact with a novel rang-free judgment algorithm for location similarity. Finally, extensive experiments conducted in a school office building show our system's good performance, and the accuracy in finding out close-contacts is more than 98%.

With the rapid development of wireless communication technology, various indoor location‐based services (ILBSs) have gradually penetrated into daily life. Although many other methods have been proposed to be applied to ILBS in the past decade, WiFi‐based positioning techniques with a wide range of infrastructure have attracted attention in the field of wireless transmission. In this survey, the authors divide WiFi‐based indoor positioning techniques into the active positioning technique and the passive positioning technique based on whether the target carries certain devices. After reviewing a large number of excellent papers in the related field, the authors make a detailed summary of these two types of positioning techniques. In addition, they also analyse the challenges and future development trends in the current technological environment.

The development of sensor technologies and smart devices has made it possible to realize real-time data acquisition of human beings. Human behavior monitoring is the process of obtaining activity information with wearables and computer technology. In this paper, we design a data preprocessing method based on the data collected by a single three-axis accelerometer. We first use Butterworth filter as low-pass filtering to remove the noise. Then, we propose a KGA algorithm to remove abnormal data and smooth them at the same time. This method uses genetic algorithm to optimize the parameters of Kalman filter. After that, we use a threshold-based method to identify falls that are harmful to the elderly. The key point of this method is to distinguish falls from people’s daily activities. According to the characteristics of human falls, we extract eigenvalues that can effectively distinguish daily activities from falls. In addition, we use cross-validation to determine the threshold of the method. The results show that in the analysis of 11 kinds of human daily activities and 15 types of falls, our method can distinguish 15 types of falls. The recognition recall rate in our method reaches 99.1%.

Carpooling is becoming a more and more significant traffic choice, because it can provide additional service options, ease traffic congestion, and reduce total vehicle exhaust emissions. Although some recommendation systems have proposed taxicab carpooling services recently, they cannot fully utilize and understand the known information and essence of carpooling. This study proposes a novel recommendation algorithm, which provides either a vacant or an occupied taxicab in response to a passenger’s request, called VOT. VOT recommends the closest vacant taxicab to passengers. Otherwise, VOT infers destinations of occupied taxicabs by similarity comparison and clustering algorithms and then recommends the occupied taxicab heading to a close destination to passengers. Using an efficient large data-processing framework, Spark, we greatly improve the efficiency of large data processing. This study evaluates VOT with a real-world dataset that contains 14747 taxicabs’ GPS data. Results show that the ratio of range (between forecasted and actual destinations) of less than 900 M can reach 90.29%. The total mileage to deliver all passengers is significantly reduced (47.84% on average). Specifically, the reduced total mileage of nonrush hours outperforms other systems by 35%. VOT and others have similar performances in actual detour ratio, even better in rush hours.

Decomposing target into several parts may improve the capability of tracking algorithm to deal with appearance variations such as occlusion and deformation. In this letter, we propose a part-based appearance model by exploiting spatial structure from parts. The model minimizes appearance and deformation cost simultaneously to predict the new position of object. Then, the optimization problem is divided into two parts. Kernelized correlation filter (KCF) is used for tracking the appearance of parts separately to speed up the proposed tracker. Meanwhile, the deformation cost is minimized by structural learning schema, which can reduce the label noise that caused by inaccuracy bounding box. Finally, minimum spanning tree and dynamic programming are employed to combine the score map of the appearance and deformation of parts, and to detect best new position of target. Experimental results on several challenge sequences show the efficiency and effectiveness of the proposed tracking algorithm.

This paper investigates the output tracking results of SISO time-delay system using the super-twisting sliding mode control approach. By a system transformation and Padé approximation, the output tracking of time-delay system is transformed into the tracking of a nonminimum phase system. Stable system centre approach is used to approximate this system by a stable one. All the tracking errors and errors from the system centre will approach to zero in a finite time by the proposed super-twisting sliding mode (SM) control. An one-link robot nonlinear example is studied to show the effectiveness of this methodology.