Time-frequency feature transform suite for deep learning-based gesture recognition using sEMG signals

摘要

Recently, deep learning methods have achieved considerable performance in gesture recognition using surface electromyographysignals. However, improving the recognition accuracy in multi-subject gesture recognition remains achallenging problem. In this study, we aimed to improve recognition performance by adding subject-specific priorknowledge to provide guidance for multi-subject gesture recognition.We proposed a time-frequency feature transformsuite (TFFT) that takes the maps generated by continuous wavelet transform (CWT) as input. The TFFT canbe connected to a neural network to obtain an end-to-end architecture. Thus, we integrated the suite into traditionalneural networks, such as convolutional neural networks and long short-term memory, to adjust the intermediatefeatures. The results of comparative experiments showed that the deep learning models with the TFFT suite basedon CWT improved the recognition performance of the original architectures without the TFFT suite in gesturerecognition tasks. Our proposed TFFT suite has promising applications in multi-subject gesture recognition andprosthetic control.