Multi-Feature Extraction EMG For Classification of Finger Movement

Many studies using EMG have been carried out, one example is the detection of hand movements with EMG. In general, research on hand movement detection uses MyoArmband as a module for recording EMG data. One of the obstacles to using the MyoArmband is that it's quite expensive and there are quite a lot of channels up to eight channels. Therefore, this study presents a comparison of the classification of five finger movements using KNN, Decision Tree, and Random Forest using only 3 EMG channels in the forearm muscles. In addition, this study also compares and combines the use of four feature extraction with Time-Domain such as RMS, VAR, MAV, and IEMG and compares the effect of using data transformation using Standardscaller on finger movement classification performance. The result of this study is that the addition of features can increase the accuracy value obtained. The best combination in this model is the RMS + MAAV + VAR feature with an accuracy of 95.15%, a precision of 95.15%, a sensitivity of 95.2%, and a specificity of 98.8%. In addition, the use of data transformation does not really affect the RF model.