Angular Velocity Estimation of Knee Joint Based on MMG Signals

Surface Electromyography (sEMG) signal is widely used as a control signal source for wearable power-assisted robots or prostheses. Most sEMG measurement electrodes need to be placed on the skin surface, and the skin should be treated accordingly, which makes the placement of acquisition equipment not convenient enough. To solve the above problems, we use multi-channel human Mechanomyography (MMG) signals to obtain human knee joint motion information, and select SENTOP angle sensor to obtain knee joint angle and angular velocity information. SVM regression model based on MMG signals for estimating knee joint angular velocity is built. In this paper, the root mean square (RMS), mean absolute value (MAV), mean power frequency (MPF), Sample entropy (SampEn) and Spearman’s correlation coefficients (SCC) of MMG signal are extracted as input of SVM regression model. Then, the prediction accuracy of SVM regression model used different features are compared. The experimental result shows that the coefficient of determination (R2) of SVM regression model reaches 0.81 ± 0.02 when all the above features are selected as input. This paper provides a method for further obtaining torque for motion control of wearable power-assisted robots with lower limbs.