Estimation of Joint Angle Using sEMG Based on WOA-SVR Algorithm

Aiming at the problem of estimating joint angle based on surface electromyography (sEMG) signals in the upper limb therapy of rehabilitation robots, a whale optimization algorithm (WOA) optimized support vector regression (SVR) is proposed to overcome the shortcomings of a single machine learning model with insufficient generalization ability. In this paper, the WOA-SVR model is established for the mapping relationship between the joint angle and sEMG signal. The inputs of the WOA-SVR model are time domain features extracted from sEMG signal, including root-mean-square and wavelength. In the experiment, two channels of sEMG signal were acquired respectively from five healthy participants, and the elbow joint angles were recorded simultaneously. The results show that, the WOA-SVR model has lower RMSE and higher R2 compared with SVR. The joint angle is successfully estimated with a RMSE error of down to 10.86(11.47±0.53). The R2 of WOA-SVR reaches 0.88(0.85±0.02) and the R2 of SVR is 0.61 on average for 5 subjects. It is found that WOA-SVR model has a good performance on estimating the joint angle. The algorithm would be used on the non-invasive and quantitative evaluation of upper-limb motor function in rehabilitation effect.