Font Size: 

The major problem in predicting elbow joint angle based on electromyography (EMG) is the non-linearity of the features. The purpose of this study is to improve the linearity of the EMG features and the accuracy in estimating the elbow joint angle based on EMG signal. To perform the proposed method, four healthy male participants were involved in this study. The EMG signal was collected from biceps using two electrode Ag/AgCl. In data collecting process, participants wore an exoskeleton frame used to synchronize the movement in flexion and extension. Eight-time domain features were investigated resulting in the best accuracy and linearity. Furthermore, to improve the accuracy and linearity of the prediction, Kalman filter was applied to optimize the EMG features. The evaluation was performed by calculating the linear regression parameters. The results (mean ± standard deviation) of the proposed method ranged between 0.721±0.145 and 0.865±0.072 for the slope, 0.899±0.075 and 0.955±0.013 for correlation and 0.815±0.127 and 0.913±0.026 for R². The improvement from non-optimized to optimized features was 57.14%, 40.30%, and 94.97% for slope, correlation, and R² respectively. These results proved the effectiveness of the proposed method to predict the elbow joint angle.

Electromyography; elbow-joint estimation; Kalman filter; time domain features