Publisher Copyright: © 2026 IEEE. EC/HE/101071179/EU//SUSTAIN EC/HE/101099491/EU//HOLDEN

Recent years have witnessed considerable efforts in repurposing ubiquitous wireless communication signals for non-contact sensing.$\bf{NearLink}$is a new-generation short-range wireless communication protocol, which is designed to address the high-quality network connectivity requirements of low power, low latency and high reliability. Given these notable advantages, NearLink has great potential for widespread application in various Internet of Things (IoT) areas. However, NearLink-based wireless sensing has not yet been explored. To bridge this gap, this work explores for the first time the sensing potential and opportunities of NearLink. Specifically, we systematically investigate the sensing capability of NearLink through answering two key questions: (1) How can NearLink's communication-oriented signals be adapted for sensing tasks? (2) How can sensing performance be enhanced under multipath interference and low-power constraints? We prototype the NearLink sensing system-NearSense, and take the respiration detection as a case study to demonstrate its effectiveness. Extensive experiments demonstrate that NearSense can achieve an average detection rate of 98% and a false alarm rate below 1.5% in case of various real-life challenging interference. We believe this work opens up new directions for the new-generation wireless sensing towards high-quality network connections.

With the rapid development of intelligent operation and maintenance of substations, higher requirements have been put forward for real-time monitoring of power equipment status. Infrared thermography, as a non-contact temperature detection means, has been widely used in the monitoring of power equipment. Based on this, this paper proposes an infrared image detection algorithm for switchgear power equipment (SCAD-DETR) based on the improved Real-Time Detection Transformer (RT-DETR) framework. First, we design a novel lightweight network architecture by combining the SMAFB (Synergistic Multi-Attention Transformer Block) and Convolutional Gated Linear Unit (CGLU) structures, which can effectively capture local and global information and model remote dependencies, thus improving the network’s ability to understand complex structures and understanding of long-distance information. Second, a multi-scale feature transformation self-attention module (MFTself Attention) is proposed, which can ensure efficient feature expression while retaining more spatial detail information, and significantly enhances the perception of regional information at the advanced feature level. Finally, we design a hybrid distillation strategy to finely optimize our model, aiming to improve the detection accuracy without increasing the computational cost. Experiments show that the proposed model achieves 99.5% mAP on switchgear equipment infrared data, which is a 5.6% improvement compared to the baseline model. In order to verify model generalizability, we further analyze and validate the model using publicly available datasets in this field, achieving 90% mAP and verifying its potential application in power equipment safety monitoring.

Long-term monitoring of human vital signs like respiration and heartbeat is crucial for the early detection of diverse diseases and overall health monitoring. Contact-free vital signs monitoring using wireless signals, particularly mmWave-based methods, has gained attention due to its sensitivity and privacy-preserving benefits. However, we observe that even minor human movements could lead to significant mutations in the signal-to-noise ratio (SNR) of the wireless signal, which cannot be explained by the commonly used model that represents the human chest as a single reflection point. These fluctuations challenge the robustness of heart rate and heart rate variability (HRV) monitoring due to the vulnerability of faint heartbeats to noise interference. To tackle this, we introduce a multi-point reflection model to understand the underlying causes of SNR fluctuations and propose a frequency diversity based algorithm to enhance sensing SNR. Our solution, Robust-Vital, was rigorously evaluated using commercial mmWave radar systems and demonstrated superior performance on long-term heart rate and heart rate variability tracking in a user study with 12 participants.

Repeated epileptic seizures impair around 65 million people worldwide and a successful prediction of seizures could significantly h elp p atients suffering from refractory epilepsy. For two dogs with yearlong intracranial electroencephalography (iEEG) recordings, we studied the influence of time series nonstationarity on the performance of seizure prediction using in-house developed machine learning algorithms. We observed a long-term evolution on the scale of weeks or months in iEEG time series that may be represented as switching between certain meta-states. To better predict impending seizures, retraining of prediction algorithms is therefore necessary and the retraining schedule should be adjusted to the change in meta-states. There is evidence that the nature of seizure-free interictal clips also changes with the transition between meta-states, which has been shown relevant for seizure prediction.