EMG Signal Recognition of Gait Pattern Using Back Propagation Neural Network for Stroke Disease Rehabilitation

Abstract

Electromyography (EMG) is the electrical activity obtained from muscles activity. Gait pattern of leg muscles will be measured and recognized by EMG signals. The EMG signal on the leg muscles is measured by six electrodes which are filtered with 0.33Hz high pass filter (HPF) and a low pass filter (LPF) for anti aliasing. Maximum frequency of EMG is 600 Hz, that sampled perfectly by Analog to Digital Converter (ADC) using 2 KHz. Artificial Neural Networks (ANN) algorithm is applied to obtain the accuracy and optimization of EMG signal. The performance of the results are investigates based on two type of human condition, first is healthy person and second is severe person. The combination of EMG measurements with ANN has gives better results compares than without an ANN model. The results showed that the measurement for healthy individuals during normal walking conditions was 4.56 volts with a frequency of 0,00582Hz; the measurement of stroke patients which walking at normal speed is 8.80 volts and the frequency is 1,231Hz. Therefore, proposed prototype that combined using ANN algorithm has increased capability of  measurement of EMG signal for normal and severe humans models.

Keywords : EMG, ANN, Gait