Design and implementation of prosthetic hand control using myoelectric signal

Abstract

Amputation is a medical procedure that is required to cut part of or all of the extremity, i.e. upper limbs or lower limbs. In the final phase of the procedure, patients have to adapt to their new condition including the use of prostheses. Nowadays, Prosthetic hand have had a lot of improvements that enable patients to do normal activities by exploiting their myoelectric signal. This study has a goal to produce prosthetic hand that can respond to patient generating myoelectric signal. Three muscle leads (2 on muscle flexor digitorum, 1 on muscle extensor digitorum) were processed by 3 channels surface electromyography (sEMG) that contain of instrument amplifier i.e. high-pass filter, rectifier, and notch filter. Myoelectric signal is processed to extraction feature and classified by artificial neural network (ANN) that had been offline-trained before and had 21 neurons input layer, 10 neurons hidden layer, and 3 neurons output layer to detect 3 hand movements, i.e. grasping, pinch, and open grasp. ANN and prosthetic hand control was embedded on Arduino Due microcontroller so that the system could be used in stand-alone and real time mode. The results of the testing from 4 research subjects shown that the hand prostheses system had success rate of 87%-91%.