Online recognition of hand EMG patterns and grasping force based on support vector machines

To solve the difficult problem of myoelectric control of a multi-DOF prosthetic hand, the research on recognition of manifold hand gesture patterns together with regression of the hand grasping force both from my electric signals was performed. Mainly based on the support vector machine (SVM) , total 18 hand gestures were discriminated by extracting the mode information from six channel surface electromyography (EMG) signals. Then the performance of the regression of the hand grasping force from EMG signals was validated under three different grasping modes. The experimental results show that the SVM methods can both effectively recognize the multi-hand gestures and estimate the hand grasping force. Combining the methods of the EMG pattern recognition with the grasping force regression, a powerful intuitive force control of a multi-DOF prosthetic hand can be realized, which can greatly improve the prosthetic hand' s control flexibility and grasping functionality.