Multi-Scale Convolution Attention Neural Network for Gesture Recognition
P H Ji, Chongli Cao, Hang Zhang, Qi Li · 2024
Surface electromyographic (sEMG) signals, as physiological indicators reflecting human body movements, hold great potential for widespread applications in medical, rehabilitation, and human-computer interaction fields. Traditional gesture recognition methods are hindered by the influence of feature selection, involving cumbersome and time-consuming feature extraction processes. In recent years, the rise of deep learning technologies has offered a promising solution to this issue. This paper introduces a multi-scale convolution attention network framework, combining Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM), along with the incorporation of attention mechanisms, to efficiently classify sEMG signals. Experimental results demonstrate significant gesture recognition performance on the Ninapro database, providing an effective solution for surface electromyographic-based gesture recognition.