Comparing EMG Signal Classification with Raw and Time Domain Feature Extracted Data

The Electromyogram (EMG) signal is a commonly used signal in biomedical engineering for studying muscle activity. This study investigates, the classification of EMG signals using raw data and Time Domain (TD)extracted data using popular Machine Learning(ML) algorithms such as Decision Tree(DT), Random Forest(RF), Naive Bayes(NB), Support Vector Machine(SVM),K-Nearest Neighbor(KNN), Logistic Regression(LR) and Linear Discriminant Analysis (LDA).Two online dataset one with 10 different finger movements, the other with 6 different wrist movements was analysed using the above mentioned algorithms. A custom made data set for 5 wrist actions was also classified. The EMG signals were collected from nine healthy individuals using a single channel with three electrodes from an EMG machine. On unprocessed data and TD data that was taken from three datasets, ML methods were used. The accuracy for the finger data, wrist movement data, and custom generated data showed that RF performed the best, with accuracy values of 0.97, 0.84, and 0.94 percent, respectively. Finally, Principal Component Analysis(PCA) was carried out, and outcomes were attained.