Human lower limb motion recognition method via surface electromyography

Abstract

Aiming at the requirements of the accuracy of human action intention recognition during the active training of lower limb rehabilitation training robot. Firstly, the mathematics model of surface EMG generation process was established and the motion perception principle with surface EMG characteristic frequency and variance was put forward. Then, the surface EMG signals from four muscles were sampled and the feature vectors were extracted. Finally, the least squares support vector machine mothed was used to establish the mapping model between feature vectors and three motions. The experimental results show that the average correct rate may reach 99%, which is 7.7% higher than the method using wavelet coefficients. It is believed that the method proposed is an efficient method.