Abstract: A novel way was presented to drive a virtual hand by using mechanomyography(MMG) signal. The MMG signal of the hand opengrip motion is acquired from the forearm, and seven kinds of timedomain features were extracted. Linear discriminant analysis was used for hand motion modes recognition. The final classification accuracy of motion modes is (95.63±2.55)%. Motion recognition results were utilized to generate proper pulses to manipulate a virtual prosthesis. The results show that the MMG signal has high accuracy of judging movements, and provides basis for prosthetic control of using MMG signal.