Optimisation and Classification of EMG signal using PSO-ANN

Abstract

Qualitative feature extraction from Electromyogram (EMG) signal has become necessary to assess the fitness of human being. Till date, various analysis tools have been employed to examine the EMG signal. Here the authors are endeavored to apply PSO-ANN based optimisation and two classification tools, namely KNN (nearest neighbor) and SVM (support vector machine) to extract features from EMG signal. EMG signal represents the signal generated by neuron from the brain, which is transmitted through the spinal cord into the body to which part is guided by the brain. The EMG signal is computed by Biopac MP45 Biomedical measurement device which is further divided into five-second segments for each activity. Unwanted EMG signal is regarded as noise and is filtered by an appropriate filter to improve the signal to noise ratio. Fourteen different time-domain and frequency domain features have been extracted for different hand movement (Weight lifting Up, Weight lifting Down, movement of Hand Gripper). Both hands are utilized for acquisition of EMG for hand grip movement. Classifier Model is used in classifying the optimised features and calculation of sensitivity, selectivity and precision of those features. From results it is evident that better accuracy is achieved for classifier KNN with respect to SVM.