COMPARISON OF TIME, FREQUENCY AND TIME- FREQUENCY DOMAIN FEATURES OF EMG SIGNALS USING MACHINE LEARNING FOR MUSCLE MOVEMENT CLASSIFICATION

• Main Author: Adzkia, Muhammad
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/76190
• Institution: Institut Teknologi Bandung

Electromyography (EMG) has been used extensively in motion recognition for rehabilitation or prosthetic control. One of the challenges in EMG-based motion recognition is the effective and accurate classification of EMG signals. In this study, we will compare the performance of feature extraction in the time domain, frequency domain and time-frequency domain with the SVM, KNN and Decision tree classifiers against EMG signals. EMG signal data was obtained from public datasets with a total of 36 subjects. The feature extraction method in the time domain will use statistical features such as Root Mean Square (RMS), mean absolute value (MAV), variance (VAR), zero crossing (ZC) and Integrated EMG (IEMG), in the frequency domain will use the mean frequency (MNF), median frequency (MDF), mean power (MNP), peak frequency (PKF) and total power (TP), while the method in the time-frequency domain will use the Wavelet transform (WT). EMG signal classification will be carried out using several classifiers including Support vector machine (SVM), k-nearest neighbor (K-NN) and decision tree (DT). The results showed that the method using feature extraction in the frequency domain is significantly better than the method in the time domain or the frequency domain in classifying EMG signals using a decision tree classifier with a higher classification accuracy of 99.26%. The results of this study can be used as a reference in selecting a more accurate and effective EMG signal classification method in motion recognition during rehabilitation or prosthetic control.