Reference position estimation for prosthetic elbow and wrist using EMG signals

Abstract

Progress in Bio robotics is opening several doors to mimic the system and purpose of human limbs, thus allowing better prostheses for amputees. This paper presents EMG based Human Machine Interface framework for estimate of reference positions of prosthetic arm for patients suffering from above elbow amputation. In the present study, a set of information about the four basic motions regarding elbow flexion/extension along with the twist of radius and ulna referred as pronation/supination has been acquired from bicep of human arm. Upon muscle activation, the numerical data set points representing the electromyographic intentions were recorded using the Thalamic Labs product i.e. Myo Armband. To analyze these acquired signals, feature selection followed with feature extraction was done in order to classify the input data extracted from human muscle. Afterwards, the performance of Artificial Neural Network (ANN) for motion classification was evaluated. Utilizing the Computer Aided design skills in SolidWorks®, a 2-DOF prosthetic arm model was developed. A mathematical model of its kinematics based on Denavit-Hardenberg convention was presented in Matlab®. Forward kinematics model were analytically verified using Matlab® with that of PeterCorke® Robotic toolbox. In order to check the physical significance of proposed work, the information of forward kinematics from Matlab® linked with that of Solid works and was simulated with reference to the EMG intentions that were captured and classified to generate the reference positions of above mentioned four motions. The classification accuracy obtained from ANN i.e. 91.9% is found significant (p<0.01) for a group of ten healthy subjects.