Continuous Joint Movements and Torques Estimation Using an Optimized State-Space EMG Model

Physiological signals, such as electromyography (EMG), have become widely exploited for human intention detection, joint movements, and torque prediction. This allows the generation of control signals for assistive or cooperative robotic devices in Human–Robot Interaction (HRI) applications such as exoskeletons. In this paper, we propose an optimized state-space EMG model for estimating joint torques and continuous movements. In order to directly estimate the movements and torques of human joints in open loop from EMG signals, the human joint forward dynamics model was integrated into a modified Hill muscle model. The model parameters were identified using a nonlinear optimization method. In addition, the EMG signal amplitude has been used to close the estimation loop in order to reject the accumulation of errors caused by integration errors and model uncertainty. In order to verify the effectiveness of the proposed approach, extensive experiments have been carried out on the elbow joint. A comparison between measured and estimated angular positions and joint torques was performed. Obtained results expressed in terms of RMSE were satisfactory.