An sEMG-Driven Shoulder Complex Motion Estimation Model Based on Glenoid Humerus Position Measurement and Muscle Synergy

The accurate quantification of human behavior is imperative for ensuring seamless human-robot interaction. Currently, surface electromyography (sEMG) decoding is the primary method for low-cost, long-term motion estimation and intent recognition. However, accurate motion measurement has not been prioritized in the development of sEMG decoding, and this limitation has created a gap between research and application. The shoulder is the most complex and flexible joint in the human body, with the highest measurement uncertainty. Thus, it has been the focus of research into posture measurement. This study proposes a motion estimation model that adequately matches the anatomy of the shoulder complex (SC). The proposed dynamic glenoid humerus (GH) position measurement method can be the foundation for constructing mathematical constraints on SC motion. Based on muscle synergy and angular decomposition of motion, the Hill muscle model with forward dynamics was extended and applied to the case of multi-muscle control of a single bone. The mean distance residual from the measured GH to the upper arm markers was 75.11% lower than that of the acromion. The error of upper arm angles estimation was 0.65 ± 0.19 and 0.14 ± 0.08 rad. Therefore, the model accurately estimated the three-dimensional motion of the SC, and the accurate tracking of the glenoid joint confirmed that the model can effectively analyze differences in human behavior.