Controlling hand robot using pattern recognition of finger movement

Abstract

The system enables to mimic the movement of the subject up to 87.5 percent accuracy. Electromyography (EMG) is one of the methods to control a robot's hand. This paper discusses the utilization of the signal EMG with a neural network algorithm to activate the fingers of the robot hand. The statistical analysis, which is the root mean square method, is used to train the patterns of the motion of the fingers of the user. Moreover, to sense the user's EMG signal, Myo armband is used. The results obtained reach 92.68 percent using 0.9 learning rate, 0.00001 error tolerance, one hidden layer with five nodes.