Liquid leakage detection is critical for product quality and user safety, yet existing methods require line-of-sight (LoS) or direct contact with the liquid, or could even introduce additional health or safety risks. In this paper, we propose SARLiquid, a novel mmWave-based through-package liquid leakage detection method. SARLiquid leverages mmWave signals to see through packaging and identify leakage by reconstructing mmWave images. We employ synthetic aperture radar (SAR) technique to enhance imaging resolution and tame multipath effects. We also develop a dedicated algorithm to calibrate the discontinuous phase in SAR imaging results, and propose a deep learning model for liquid leakage detection and liquid identification. We implement SARLiquid and evaluate it across a wide range of scenarios. Results show that SARLiquid achieves average accuracies above 93% for both liquid leakage detection and liquid identification, 14.70% and 9.98% higher than the baselines, respectively.

Human Action Recognition (HAR) offers a wide range of applications, including smart home, smart health, entertainment, security, and surveillance. Traditional vision-based HAR systems face significant limitations due to privacy concerns, lighting dependency, and poor performance in complex environments. Millimeter-wave radar-based activity recognition systems have attracted considerable attention due to their superior sensing capabilities, device-free deployment, privacy preservation, and robustness to environmental variations. However, existing approaches struggle with the inherent sparsity and noise in mmWave radar data, particularly for diverse activity categories spanning from full-body movements to subtle localized gestures. This study proposes mm-ARnet, a comprehensive millimeter-wave point-cloud-based framework for recognizing 16 diverse human activities across three distinct behavioral categories: full-body movements, posture transitions, and localized body movements. Our approach leverages 4D point cloud sequences and introduces a multi-frame fusion with stochastic sampling strategy to enhance point cloud density and mitigate sparsity effects. The core innovation lies in our lightweight TCN+Bi-LSTM temporal modeling pipeline integrated with a novel Temporal Pattern Attention (TPA) mechanism. Extensive experiments conducted across three real-world scenarios with 10 participants demonstrate that mm-ARnet achieves 97.42% accuracy, outperforming state-of-the-art methods while maintaining superior temporal performance.

Human action recognition in classrooms has recently become a research hotspot. Traditional solutions usually rely on sensors or computer vision methods. However, these methods have some disadvantages, such as difficulty in deployment, susceptibility to ambient light, and privacy and security issues. This paper proposes RFAR, a contactless method for classroom action recognition. This method utilizes an RFID tag placed on the desktop to capture various actions and subsequently evaluate the student's learning status. To enhance the reliability of singletag identification, fused data consisting of two or three types of data sequences (RSSI, phase, and Doppler shift) are incorporated. Furthermore, a dynamic antenna system is utilized to identify the optimal angle for tag-antenna alignment. Notably, the single-tagper-person design eliminates severe interference among multiple tags and simplifies device deployment in multi-person scenarios. This method is proposed based on COTS RFID devices and shows high robustness across different environments and equipment. Experimental results show a recognition accuracy of 93.9% in single-person scenarios and 81.5% in five-person scenarios.

Facial expression recognition technology plays a significant role in advancing the intelligence and personalization of human-computer interaction. Although ultrasonic-based expression recognition methods already exist, most rely on phase shifts or Doppler effects, which suffer from insufficient resolution to accurately capture subtle facial expression variations. To address this issue, this paper proposes an innovative facial expression recognition system, SonicFER, which utilizes smartphones to emit ultrasonic waves and receive echoes, achieving fine-grained perception of facial expressions through real-time monitoring and analysis of channel impulse response (CIR). To effectively detect facial expression movements while eliminating static interference and minor motion artifacts, this paper employs a differential operation combined with variance calculation, along with setting appropriate thresholds for filtering. Through rigorous experimental evaluation, SonicFER achieves a high accuracy of 91.2% in recognizing six common facial expressions and outperforms state-of-the-art technologies across various real-world scenarios.

Existing floor localization methods are plagued by low accuracy, high algorithmic complexity, dense node deployment, susceptibility to environmental factors, and the inability to track trajectories. This paper introduces a localization method designed to address the challenges of multi-floor environments, leveraging LoRa technology. The approach involves deploying LoRa vertical positioning devices and establishing offline and threshold fingerprint databases. To enhance localization accuracy, it combines Time-of-Flight (TOF) ranging values (referred to as "RANGE" in this paper) with Received Signal Strength Indicator (RSSI) values, referred to as "RSSI-RANGE". Subsequently, a multi-floor determination is achieved using the RSSI-RANGE floor determination algorithm and a range-based signal source autonomous switching mechanism. The fingerprinting technique is then employed for trajectory recognition. Comprehensive vertical information is obtained by combining floor determination and trajectory award. Gaussian filtering is utilized for fingerprint preprocessing to eliminate gross errors. The particle swarm optimization algorithm is employed to fine-tune the hyperparameters of the random forest algorithm following noise reduction. Using the random forest algorithm, optimal RSSI-RANGE values are derived, and the offline fingerprint database is established by applying Kriging interpolation. Localization is then achieved in the concluding online recognition phase. Empirical findings illustrate the system's high floor accuracy rate of 97.8%, achieving high determination accuracy and comprehensive floor localization when combined with trajectory recognition.

Sign language plays a vital role in communication and learning for individuals with hearing and speech disabilities, serving as a common language for the deaf. Current state-of-the-art sign language recognition methods primarily rely on computer vision techniques, but they have certain limitations, including susceptibility to light interference and privacy concerns. Ubiquitous acoustic sensing provides new possibilities for sign language recognition, leveraging its high resistance to interference and cost effectiveness. However, existing methods face challenges in achieving satisfactory results due to environmental interference and the complexity of sign language recognition contexts. In this work, we propose EarHear, a robust contactless Chinese Sign Language Recognition and translation system. EarHear adopts a differential-Doppler data preprocessing method to cleverly mitigate the interference caused by the environment. To further identify differences in the morphology, speed, and direction of sign language actions and distinguish similar gestures, we propose the vision transformer for sign language recognition, which is able to model the context dependence of long-range features and output indeterminate long sign language sequences using an attention mechanism. As a result, computational speed and recognition accuracy are improved. Moreover, we explore a large-scale language-model-based sign language translation, which enables sign language recognition results to follow natural language standards, thus realizing a true sense of sign language recognition. The evaluation results based on 15 Chinese sentences show that our system achieves an average recognition rate of 93.38% and a BLEU-1 score of 80.73% for sign language translation, reaching the most advanced level in terms of accuracy and robustness.

Millimeter-wave integrated access and backhaul network (mmWave IABN) is a cost-effective paradigm to accommodate the ever-increasing demands of IoT devices, but its sensitivity to the obstructions is an obstacle in the practical commercial applications. Reconfigurable intelligent surface (RIS) is an innovation enabler to proactively manipulate the ambient environment in a programmable manner for the overall performance enhancement. In this paper, the resource allocation of the RIS-assisted mmWave IABN is conceived and designed to maximize the end-to-end throughput. Firstly, the end-to-end throughput maximization oriented resource allocation problem including transmission power controlling, phase-shift manipulation, and bandwidth allocation is formulated as a non-linear and non-convex programming problem, which is intractable to search an optimal solution in polynomial time. Secondly, the formulated original problem is equivalently decomposed into two subproblems to obtain a sub-optimal solution, i.e., the joint optimization subproblem of transmission power controlling at the users and phase-shift manipulation at the RIS, and the bandwidth allocation subproblem between the access part and the backhaul part. Thirdly, a decomposition iteration based resource allocation mechanism (DIRAM) is proposed, and the DIRAM is composed of two phases, which is the alternative iteration based optimization scheme of power controlling and phase-shift manipulation (AIOS-PCPM) to obtain the optimal solution of the first subproblem, and the piecewise statistical based bandwidth allocation (PSBA) scheme to obtain the optimal solution of the second subproblem. Finally, the properties of the proposed DIRAM are evaluated through abundant of simulation comparisons with other baselines, where the superiorities of the proposed DIRAM are verified in terms of spectral efficiency and end-to-end throughput.

Fatigue driving is the leading cause of severe traffic accidents, which is considered as an important point of the research. Although a precise definition of fatigue is lacking, it is possible to detect the physiological characteristics of the human body to determine whether a person is fatigued, such as head shaking, yawning, and a significant drop in breathing. In our study, fatigue actions were collected first, and then the different micro-Doppler characteristics produced by human activity were used to classify and recognize the fatigue action using the fine-tuning convolution neural network (FT-CNN) model. The collected signals in the breathing mode were preprocessed to judge whether the person was fatigued according to the estimated value of the respiratory rate. Data in different environments were collected to verify the proposed method. Our results showed that the accuracy of fatigue detection can reach 91.8% in the laboratory environment and 87.3% in realistic scenarios.

One of the key techniques for future wireless network is full-duplex-enabled millimeter wave integrated access and backhaul network underlaying device-to-device communication, which is a 3GPP-inspired comprehensive paradigm for higher spectral efficiency and lower latency. However, the multi-user interference (MUI) and residual self-interference (RSI) become the major bottleneck before the commercial application of the system. To this end, we investigate the sub-channel allocation problem for this networking paradigm. To maximize the overall achievable rate under the considerations of MUI and RSI, the sub-channel allocation problem is firstly formulated as an integer nonlinear programming problem, which is intractable to search an optimal solution in polynomial time. Secondly, a coalition formation based sub-channel allocation (CFSA) algorithm is proposed, where the final partition of the sub-channel coalition is iteratively formed by the concurrent link players according to the two defined switching criterions. Thirdly, the properties of the proposed CFSA algorithm are analyzed from the perspectives of Nash stability and uniform convergence. Fourthly, the proposed CFSA algorithm is compared with other reference algorithms through abundant simulations, and superiorities including effectiveness, convergence and sub-optimality of the proposed CFSA algorithm are demonstrated through the kernel indicators.

Coughing and sneezing are important routes of virus transmission. Droplets carrying the virus enter the air and spread rapidly, increasing the spread of the disease. Therefore, how to accurately detect coughing and sneezing behaviors in a timely manner so as to effectively warn the spread of the virus has become an urgent problem. To solve this problem, we designs a coughing and sneezing detection scheme for indoor people on commercial wireless devices. First, the Doppler shift feature image caused by the action is segmented using a clustering algorithm, which reduces the computational overhead of the system. Then, the HOG features of the segmented image are extracted and input to the two-dimensional SOM network for action classification and recognition, which effectively improves the detection accuracy of target actions. Finally, a dataset consisting of real coughing and sneezing actions is constructed and open-sourced in this paper. The performance of this solution was tested and analyzed in several dimensions under two typical application scenarios. The results show the robustness of this scheme and the accuracy up to 93.1% in real-world scenarios. Our solution offers a new technology and method for disease prevention detection.

In recent years, the researchers have witnessed the important role of air gesture recognition in human-computer interactive (HCI), smart home, and virtual reality (VR). The traditional air gesture recognition method mainly depends on external equipment (such as special sensors and cameras) whose costs are high and also with a limited application scene. In this paper, we attempt to utilize channel state information (CSI) derived from a WLAN physical layer, a Wi-Fibased air gesture recognition system, namely, WiNum, which solves the problems of users’ privacy and energy consumption compared with the approaches using wearable sensors and depth cameras. In the process of recognizing the WiNum method, the collected raw data of CSI should be screened, among which can reflect the gesture motion. Meanwhile, the screened data should be preprocessed by noise reduction and linear transformation. After preprocessing, the joint of amplitude information and phase information is extracted, to match and recognize different air gestures by using the S-DTW algorithm which combines dynamic time warping algorithm (DTW) and support vector machine (SVM) properties. Comprehensive experiments demonstrate that under two different indoor scenes, WiNum can achieve higher recognition accuracy for air number gestures; the average recognition accuracy of each motion reached more than 93%, in order to achieve effective recognition of air gestures.