Study of Myoelectric Prostheses Based on Improved LS-SVM

Least Squares Support Vector Machine (LS-SVM) has many pattern recognition fields. Considering the non-steady character of electromyography signal, wavelet transform is employed to analyze electromyography on the basis of acquired signals. The singular value decomposition of a wavelet coefficient matrix is adopted to extract features of surface electromyography and the Least Squares Support Vector Machine algorithm is utilized to implement the multi-motion pattern classification of surface electromyography including hand opening, hand closing, wrist supination, wrist pronation. The experiment shows the method has a fast running speed, high discrimination rate and good robust, so it has a great potential in the area of bionic man-machine systems such as using electromyography signal to control powered prosthesis.