Gesture recognition of sEMG signal based on GASF-LDA feature enhancement and adaptive ABC optimized SVM

Abstract

The surface EMG signal (sEMG) is the potential signal produced by human muscle movement, which is closely related to the movement pattern of the limb and widely used in the field of gesture recognition. However, most of existing methods suffer from insufficient adaptability and low recognition accuracy, in order to achieve rapid and accurate gesture recognition from sEMG signals, a novel sEMG-based gesture recognition method is proposed involving feature enhancement based on gramian angular summation field-linear discriminant analysis (GASF-LDA) and improved support vector machine. Specially, feature enhancement is implemented for the extracted features by GASF-LDA algorithm. Additionally, an adaptive principle based on the artificial bee colony algorithm is proposed to optimize the support vector machine for sEMG-based gesture recognition. As a result, for the eight-classification study of sEMG signals, the combination of feature enhancement and classifier optimization achieved an average accuracy of 97.54 ± 1.03 %. The experimental results show that the method has high classification accuracy and stability.