Whole-Field Action Sensing via Wearable Single-Channel EMG Sensors and Resource-Efficient Motion Network

The proliferation of multi-person sports and collaborative training paradigms has underscored the necessity for comprehensive whole-field motion data integration. However, existing vision/WiFi-based approaches suffer from occlusion-induced tracking errors and interference-driven signal instability, compromising multi-person activity recognition performance. Notably, Electromyography (EMG) recognition, pivotal in Wearable Human Activity Recognition (WHAR) for muscle activity analysis and motion intent decoding, still struggles to balance performance and real-time efficiency in multi-person scenarios. Unlike channel-expansion approaches, we propose a wireless wearable Single-Dimensional Sparse EMG (2SEMG) Sensor for efficient personal sampling. These motion-unaffected sensors leverage the proposed lightweight One-Dimensional Motion Network (OMONet) to facilitate whole-field motion sensing. Experimental results indicate that the OMONet achieves state-of-the-art performance and efficiency in EMG and other physiological signal recognition, with its core components validated through ablation studies. The motion detection results of two representative badminton matches further validate the performance, robustness, and real-time efficiency of the whole-field motion sensing network based on 2SEMG Sensors and OMONet in real-world multi-person sports.