sEMG based movement quantitative estimation of joins using SVM method with gravitational search algorithm

Abstract

Recently the surface electromyogram signal (sEMG) based motion estimation developed rapidly, which focus on intention recognition but the other information of motion is not concerned. This paper proposed a sEMG based quantitative analysis method to estimate movement of human joints, which was used to control the upper limb rehabilitation robot system by participant's own arm. The quantitative model was established utilizing support vector machine (SVM). In order to improve the fitting accuracy and generalization ability of the support vector machine model, an algorithm for the SVM parameter optimization was proposed based on the gravitational search algorithm. The simulated experiments show that the SVM regression model based on the gravitational search algorithm has a high accuracy and strong generalization ability. Initial online experiments on rehabilitation robot controlled by a healthy participant demonstrated that the sEMG based control method using the proposed method was feasible.