Emotion stimuli-based surface electromyography signal classification employing Markov transition field and deep neural networks

Surface electromyography (sEMG) has been widely used in clinical medicine, rehabilitation medicine, and intelligent robots. Currently, sEMG signal classification methods promoted the development and industrialization of sEMG control bionic prostheses. Emotion recognition using sEMG signal is crucial in human–computer interaction (HCI) and becoming a research hotspot. While the high rate of emotion recognition is still the key issue for the emotion applications. Employing sEMG to study emotion classification can improve the recognition rate and eliminate subjective interference. In this research, the Markov transition field (MTF) method was applied to convert sEMG signals to images; and this crucial converting process makes convolutional neural networks adopting the input resource. A 69-INPUT-6 -OUTPUT primary deep neural network was constructed for classifying the human emotion states under emotion stimuli experiment. The MTF-based deep neural network (MTF-DNN) for classifying sEMG signals was developed and validated subsequently. The result showed that the high effectiveness of the proposed classification model. The proposed MTFDNN performs high efficacy in the indices of classification of Ac (0.9102), Pr (0.1867), and Fm (0.9089) by comparing with different classification models.