A Novel Estimation Approach of sEMG-based Joint Movements via RBF Neural Network

Abstract

In these years, the research of rehabilitation robot has been more and more extensive, among which the key lies in the intention recognition of human body. When the human body moves, the muscles contract will produce certain electric signals, called surface electromyography (sEMG) signals. The sEMG signals can be utilized to estimate human movement intention. The radial basis function (RBF) neural network is adopted to predict the joint angle of healthy people in this paper. The subject sit in a chair and perform leg stretching. Firstly, the sEMG signals of three muscles, namely, rectus femoris (RF), lateral femoral muscle (LF) and extensor halluces (EH) longus are collected through Biopac system, when the subject do exercise. And then the processed signals are used as the input of the network to estimate the three joint angles of human lower limbs through the network learning and training. Root-mean-square (RMS) error is used as the criterion for evaluating the model performance. Through MATLAB simulation experiment, it can be verified that RBF network can effectively estimate the joint angle of human body, among which the RMS error of hip joint is 1.02360, the RMS error of knee joint is 8.07520, and the RMS error of ankle joint is 11.06330 Therefore, the movement intention of human body can be effectively estimated through RBF neural network. In other words, using sEMG signals as input, the movement intention of the lower limbs of human can be estimated via the RBF neural network. Furthermore, the method can be generalized to the rehabilitation robot and auxiliary robot.