Hand gesture recognition based on sEMG signals using Support Vector Machines

Abstract

This paper demonstrates the application of electromyography (EMG) signals for controlling home devices. To achieve this we have used an armband called MYO® that has an array of eight sEMG sensors around the forearm. We have studied 15 different hand gestures to create a dictionary of gesture control. We have achieved gesture recognition using Support Vector Machines (SVMs) as a classification method. We tested different types of kernels (radial, polynomial and sigmoid) to achieve the optimum conditions for gesture learning and recognition as well as an accurate determination of these movements. Furthermore, to show the effectiveness and applicability of the results, a Gesture Control System has been implemented as an embedded system. The system also enable Bluetooth communication with the armband, send gesture control commands to household devices using iR protocol.