Smoothing Post-Processing for Continuous sEMG-Based Joint Angle Estimation

Surface electromyogram (sEMG) signals have attracted widespead attention from numerous researchers in the fields of intelligent interaction and rehabilitation due to their portability, non-invasiveness, and ability to be generated prior to movements. In this paper, we propose a smoothing post-processing method specifically designed for the sEMG-based bilateral upper limb rehabilitation system. Aiming at the trade-off between the volatility of sEMG prediction and the smoothness of control encountered during the continuous estimation of the elbow joint angles, we introduce the higuchi fractal dimension (HFD). By calculating the HFD of sEMG within adjacent sliding windows, the smoothing factor in exponential moving average (EMA) algorithm is dynamically adjusted, thereby achieving adaptive smoothing of the sEMG prediction results. Experimental studies conducted on three subjects demonstrate that the proposed method can effectively reduce the fluctuation amplitude of sEMG prediction values without significantly compromising prediction accuracy, and notably enhances the stability and reliability of the prediction results.