A classification of EMG signal from masseter and buccinators, muscles to control the directional movement of power-assisted wheelchair

Abstract

There are many people who cannot control the movement of their upper or lower limbs. Also, there are many people affected with some form of paralysis, suffering from a spinal cord injury, and many elderly people are unable to control their upper and lower limbs. Therefore, it is necessary to provide them with an alternative control device that can help them to achieve some mobility independence, where the wheelchair is very important for these people to help them in their daily lives for moving from one place to another in a comfortable manner. The main objective of this research work is to control the movements of the wheelchair in five directions (forward, reverse, stop, left and right), using signals from the masseter and buccinators muscles as control signals. Then extracted the features of the autoregressive model, waveform length, mean absolute value and root mean square, and then classify them by using a K-nearest neighbor classifier and linear discriminant analysis to choose the better result of the classification and utilize it as a control signals for the wheelchair movement in offline method. The result of classification shows that the accuracy of the K-nearest neighbor classifier is very higher compared with the linear discriminant analysis classifier, where the highest rate of accuracy was 98.88% when using the KNN classifier with the AR model 4-order.