FEATURES COMBINATION ON EMG SIGNALS FOR CLASSIFICATION OF FINGER MOVEMENTS

• Main Author: Arland, Firzal
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/76189
• Institution: Institut Teknologi Bandung

Currently in Indonesia technology in the field of robotics and AI is increasingly advanced and rapid. Quoted from the KEMENKES website, robotic surgery technology has started running from the Indonesian Robotic Surgery Center since 2021. Technological developments in the field of rehabilitation such as the detection of muscle strength in stroke patients using EMG. EMG is a measurement technique used to record the electrical activity generated by the muscles of the human body. The use of EMG has also been widely studied in the medical field. One of the studies using EMG is the detection of finger movements using the MyoArmband. MyoArmband is an EMG module that will detect the user's muscle contractions. The MyoArmband is shaped like a bracelet and has eight EMG channels. However, research using MyoArmband is an obstacle, because the price is quite expensive and requires a fairly complex system. Therefore, in this study, we propose to use MyoWare as many as three channels which are useful for reading hand muscle activity. Three MyoWare channels will be placed on the forearm area to detect finger movements. The value obtained will be stored in the form of CSV data. The stored data will go through the feature extraction stage to get its Fitur. Feature extraction is carried out such as RMS, MAV, VAR, and IEMG in the time domain. These Fitur will later be combined so that the classification performance can be better. In addition, this study also compares the use of data transformation in the form of a standard scaler which will be processed using several machine learning classification methods including k- Nearest Neighbor (KNN), Decision Tree (DT), and Random Forest (RF). The results of this study are that by combining Fitur for classification of finger movements, it can increase the accuracy of the classification performance. The feature combination that gets the highest accuracy value is the RMS+MAV+VAR combination using a standard scaler. The accuracy value obtained reached 95.19% with a precision value of 95.15%, a sensitivity of 95.2% and a specificity of 98.8%.