Tracking control of real-time force for prosthetic hand based on EMG

Aiming at such problems as low level of flexibility and intellectualization existing in commercial prosthetic hand, the study on a five-fingered prosthetic hand with moment/position perception and its control method was performed. The electromyography (EMG) signal of body forearm was taken as a control source, and the fingertip normal force was tested with a force sensor. A prediction model for fingertip force was established with a support vector regression (SVR) machine. The parameters of the proposed model were determined with a grid search method. And thus, the training time is shortened, and the generalization ability of the model gets enhanced. The real-time tracking of fingertip force was realized through combining the online prediction and fuzzy PID control method with a second-order differential tracker. The proposed control method has quick response speed, small ultra regulation and strong robustness, and thus, the control level for prosthetic hand gets improved.