sEMG Based Continuous Estimation of Wrist Joint Angle using BP Neural Network

Abstract

The estimation of joint angle based on sEMG is one of the key technologies in bioelectric processing, which could be used for medical rehabilitation robots, intelligent prosthetics, industrial mechanical arms and so on. However, the continuous estimation of wrist joint angle based on sEMG is rare. This paper implements continuous estimation of wrist joint angle based on sEMG. The sEMG and the motion data of wrist joint were acquired firstly. Then six kinds of sEMG features were extracted and the actual angle of wrist joint was calculated. Thirdly, the BP neural network was established to analysis the relationship between sEMG and the motion angle of wrist joint. Finally, the verification experiment was performed, and results show that the MAV of sEMG could acquire the best estimation effect and the mean linear correlation coefficient between estimation angle and actual wrist joint angle reached 0.98, which proves the proposed method is effective. So the research in this paper makes up for the lake of wrist joint angle decoding, which increases the flexibility of robots operational and expand the application filed of decoding joint angle based on sEMG.