Feature Extraction and Classification Between Control and Parkinson’s Using EMG Signal

Abstract

The main objective of the proposed work is to classify and differentiate Parkinson’s disease from other neuromuscular disease with the help of Electromyogram (EMG) signal. An Electromyogram signal detects the electric potential generated by muscle cells when these muscle cells contract or relax. However, these electromyography signal itself failed to differentiate between these neuromuscular diseases as their symptoms are almost the same. Therefore, certain features were examined and studied like average distance peaks, discrete wavelet functions, entropy, band power, peak-magnitude-to-RMS ratio, mean complex cepstrum and maximum value of single-sided amplitude, etc. These features were extracted, and with these features, we can differentiate between these neuromuscular diseases including Parkinson’s disease. Two classifiers were used for detection and classification of Parkinson’s, they were Decision Tree and Naive Bayes. However, the accuracy in Decision Tree was found out to be 88.38%, while the accuracy in Naive Bayes was found out to be 54.07%.