Feature Extraction from sEMG of Forearm Muscles, Performance Analysis of Neural Networks and Support Vector Machines for Movement Classification

The propose of this work is to extract different features from surface EMG signals of forearm muscles such as MAV, RMS, NZC, VAR, STD, PSD, and EOF's. Signals are acquired through 8 channels from "Myo Armband" sensor that is placed in the forearm of the human being. Then, identification and classification of 5 types of movements are done, including open hand, closed hand, hand flexed inwards, out and relax position. Classification of the movement is performed through machine learning and data mining techniques, using two methods such as Feedforward Neural Networks and Support Vector Machines. Finally, an analysis is done to identify which features extracted from the sEMG signals and which classification method present the best results.