A Simplified Convolutional Block Attention Module for Robust Hand Gesture Recognition with High Density Surface Electromyography
Bangyan Niu, Junwei Li, Yifan Wang · 2025
High-density surface electromyography (HD-sEMG) arrays have become increasingly prevalent with advances in flexible electronics, providing abundant neuromuscular signals. While these high-density electrode arrays enhance spatial resolution, they also introduce challenges in pattern recognition, as the overwhelming amount of redundant and noisy signals may degrade classification performance. To address this issue, we propose a simplified Convolutional Block Attention Module (sCBAM) for robust and high-accuracy hand gesture recognition. By sequentially applying a channel attention module (CAM) and a spatial attention module (SAM), the feature representation can be effectively refined, emphasizing the most relevant information while suppressing noise and redundancy. The output attention map builds the relationship between activated muscle units and identified hand gesture, making the process of recognition no longer a total black box. Besides, the proposed sCBAM is highly lightweight and can be plugged into existing CNN architecture for better feature representation. Experiments are implemented based on an open access HD-sEMG dataset Hyser. The comparison results and ablation study confirm the effectiveness of the proposed sCBAM module.