Study of Myoelectric Prostheses Based on Improved LS-SVM and Fuzzy Control

The sticking point in studying multi-freedom myoelectric prostheses is based on multi-motion pattern recognition of surface electromyography, therefore, in this paper, a method that takes those singular eigen values of wavelet coefficients as the eigenvector of improved least squares support vector machine (LS-SVM) is presented to discriminate the motion pattern. Considering the non-steady character of electromyography signal, wavelet transform is employed to analyse electromyography on the basis of acquired signals that have been pre-processed earlier, consequently 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 recognition of surface electromyography. Then a fuzzy controller is designed specially to control the adjustment of myoelectric prosthetic hand´s movement, which can make the system of myoelectric prosthetic hand grasp object stably. Experimental results indicate that above method has a fast running speed, high discrimination rate and good robust, it can increase the correct ratio of movement pattern recognition and decrease the possible damage to grasped objects, so it has a great potential in the area of bionic man-machine systems such as using electromyography signal to control powered prosthesis.