Effect of shoulder angle variation on sEMG-based elbow joint angle estimation

For the decade now, surface electromyogram (sEMG) signal has been extensively applied in joint angle estimation to control the prostheses and exoskeleton systems. However, the sEMG signal patterns can be severely affected by shoulder angle variations, which restricts its applications to a practical use. In our study, we evaluate the effect of shoulder angle variations on elbow angle estimation performance. This adverse effect increases mean root mean square (RMS) error by in our experiment. Then, four estimation methods are proposed to solve this problem: (1) using a training set including all shoulder angles' training data to train model; (2) adding two shoulder muscles' sEMG as additional inputs; (3) a two-step method using arm muscles' sEMG and two shoulder muscles' sEMG; and (4) a two-step method using arm muscles' sEMG and measured shoulder angle value by a motion sensor. 13 subjects are employed in this study. The experimental results demonstrate that the mean RMS error is reduced from to in method one, in method two, in method three, and in method four, respectively. These results show that our methods are effective to eliminate the adverse effect of shoulder angle variations and achieve a better elbow angle estimation performance. Furthermore, this study is helpful to develop a natural and stable control system for prostheses and exoskeleton systems.