Joint Motion Prediction Based on sEMG Third-order Cumulant Feature

Abstract

In order to solve the problem that the surface electromyography signal is not accurate enough for the quantitative identification of the human motion angle, this paper used PCA-based third-order cumulant analysis method to extract sEMG signal features, and used this features as input for a single-layer BP neutal network to predict the joint angle of the human body. In the identification of shoulder joint motion, the root mean square error of the prediction result was 5.19, and the correlation coefficient was 0.95, which is obviously superior to the AR model method and the RMS method.