Sensor-assisted EMG data recording system

Abstract

We present a sensor-assisted electromyography (EMG) data recording system in this paper. The system consists of two types of sensors: (1) hand gestures acquired by using 5DT data glove and (2) EMG sensor acquired by using MYO armband. The additional information from 5DT data glove can help in identifying hand gestures resulting in more convenient, fast, and precise EMG data collection procedures compared with the conventional system using EMG information solely. The performance of the proposed system was validated by classifying 3 hand gestures consisting of rest hand, open hand, and close hand. Six features are extracted from 8-channel EMG signals from MYO armband. While principal component analysis (PCA) was employed as a dimensionality reduction technique, support vector machine (SVM) was used as a classifier. The window size 100 samples with overlapping 50% gives the maximum average classification accuracy at 91.28%. On the other hand, the window size 100 samples with overlapping 75% gives the maximum rate of analysis at 120 predictions per minute.