An Approach for sEMG-based Gesture Recognition Using Continuous Wavelet Transform and AlexNet Convolutional Neural Network

To improve the classification accuracy of gesture recognition based on sEMG signals, an approach for applying continuous wavelet transform (CWT) and AlexNet convolutional neural network is proposed in the paper. Based on the principle and feature analysis of gesture-based sEMG signals, the CWT-AlexNet model is set up for gesture recognition. Firstly, continuous wavelet transform is exploited to extract time-frequency domain features and create scalograms from the sEMG signals. Then, the two-dimensional time-frequency images are fed to the AlexNet network as feature images, achieving classification of different hand gestures. Rami Khushaba EMG repository comprised 10 gestures of 8 healthy subjects is applied to validate the model performance. A comparative study of classifying gestures of 4 single subjects and multi-subjects is implemented in the paper. The results show that the proposed approach perform well on gesture recognition, giving offline accuracy above 99.9% for different subjects datasets.