Continuous estimation of hand's joint angles from sEMG using wavelet-based features and SVR

Abstract

Developing robust hand kinematic estimation mechanisms is considered an essential requirement to enhance the quality of life for amputees. These robust control mechanisms enable to control robotic hands in a way that can mimic the human hand functions. In this paper, we propose a surface electromyography (sEMG)-based approach for continuous estimation of wrist and fingers' joint angles. The proposed approach utilizes the discrete wavelet transform (DWT) to construct a time-frequency representation of the sEMG signals. Then, using the time-frequency representation, a set of time-frequency features are extracted. In order to estimate the wrist and fingers' joint angles, we utilize the extracted time-frequency features to train a set of support vector regression (SVR) models. Evaluation results of the proposed approach, using the NinaPro database, demonstrate the efficiency of the approach in providing a feasible method towards accurately estimating wrist and fingers' joint angles from the sEMG signals.