Recognition of continious arm movement based on electromyography data

Abstract

Currently, neural-computer interfaces require expensive hardware, which is not available for most researchers, while EMG sensors are cheap, affordable, and quite robust. That makes them an attractive option for a wide class of devices, like prostheses, game devices, or exoskeletons. So reliable and accurate methods of EMG data recognition and interpretation are required. While most of the popular methods of EMG data analysis include only distinct gesture recognition, in this thesis we try to implement the system, which recognizes continuous motion on the example of arm movement and end effector (palm) pose estimation. This thesis goal is to prove that this kind of estimation is possible by creating a system that will estimate arm position in 3d space.