Gesture Recognition System Based on Surface Electromyography Signals of the Forearm

This paper designs and implements a gesture recognition system based on surface electromyography (sEMG) signals of the forearm, covering both the hardware acquisition device and innovative algorithms. The hardware system is centered on the nRF52832 microcontroller, combined with high-precision amplification, filtering, and A/D conversion modules. It can collect multi-channel sEMG signals in real-time and transmit them to the host computer via Bluetooth. In view of the characteristics of electromyography signals being weak and easily disturbed by noise, the system adopts technologies such as differential amplification, right-leg drive, and π-type filtering, which effectively improve the signal accuracy and anti-interference ability. In terms of gesture recognition algorithms, an end-to-end recurrent neural network (RNN) framework is proposed. The short-time Fourier transform (STFT) is used to generate time-frequency diagrams, and a two-layer RNN is adopted to identify the muscle activation areas and classify gestures respectively. Experimental results show that compared with methods such as CNN-LSTM, Temponet, and ALR-CNN, the algorithm in this paper has significant advantages in both accuracy and operational efficiency. In the task of classifying seven types of hand movements, this method achieved a classification accuracy of 89.45%, an F1 score of 95.67%, and significantly reduced computational complexity, making it suitable for real-time gesture recognition applications on resource-constrained devices. The research in this paper provides innovative ideas for the efficient processing and application expansion of sEMG signals.