The applications of present commercially-available myoelectric prostheses are limited due to the difficulty in use and the long training time. The existing control methods can not realize natural controls of the prostheses. In this paper, a control system of multifunctional myoelectrical prostheses which contains a myoelectric-prosthesis controller based on a micro control unit and a multifunction prosthesis clinical training software, was designed and developed using a high-performance microprocessor. The pattern recognition of the electromyogram (EMG) was employed to determine the motion classes. Since the microprocessor is wearable, the system proposed in this study could possibly achieve flexible and natural controls of multifunctional myoelectric prostheses.