An Efficient Feature Extraction Technique and Novel Normalization Method to Improve EMG Signal Classification

Electromyography (EMG) is one of the medical diagnostic techniques that is widely used to investigate muscular health status. Medical and paramedical professional’s record EMG signals as myoelectric signals that need some processing to improve their apt analysis and interpretation. In the present paper, the authors have integrated feature extraction and normalization step to improve the overall EMG signal classification. The process involves 12 feature extraction matrices that are used for the normalization of the recorded signals. The processed signals are fed to a binary classifier, Support Vector Machine (SVM) for categorization into pain and normal classes. Performance analysis is performed using 100 simulation rounds and the observations are summarized in terms of precision, sensitivity, specificity, and accuracy of classification of EMG signals. The simulation analysis demonstrated that with the integration of feature extraction and normalization step, an improvement of 3% is observed in the EMG signal classification accuracy.