sEMG-Based Gesture Recognition Using Temporal Convolutional Network

Surface electromyography (sEMG) signals can rapidly and accurately reflect human movement intentions, and gesture recognition based on sEMG signals has been extensively studied in the field of human-machine interaction (HMI) for rehabilitation robots. Due to the complexity and nonlinear relationship of the signals, to achieve more precise gesture classification, this paper proposes a sEMG gesture recognition algorithm based on Temporal Convolutional Networks (TCN) that fully explores signal characteristics. The study investigates the classification of electromyographic signals during different gestures performed by subjects. After noise reduction and filtering of the raw signals, four eigenvalue sequences are extracted as the input signals for the TCN model, with gesture classification labels serving as the model output. The accuracy of the model is evaluated using test data. The experimental results show that the gesture recognition aligns with the actual categories, with an approximate classification accuracy of 95.5%, validating the accuracy of the TCN model in gesture recognition. This research has potential applications in optimizing exoskeletal robots, controlling active prosthetics, and regulating assistive robots.