Abstract: As most electromyography prosthesis controller systems are based on a threshold to control hand’s opening and closing, with poor operational flexibility, it is proposed to use ARM in the system. A STM32 ARM core processor is used as the main chip. It collects the flxor carpi ulnaris and flexor carpi radialis electomyography signals with 2 A/D signal converters, and extractes 4 kinds of characteristic values both in the time and frequency domains. By using a BP neural network classification algorithm, the system realizes real-time online identification of 5 kinds of palm action modes. Experimental results show that the system’s online recognition rate for the 5 actions is up to 97%, meeting the real-time requirements of prosthetic hand control.

Key words: BP neural network, electromyography, feature extraction, prosthetic real-time control, ARM