Multibody kinematics optimization for motion reconstruction of the human upper extremity using potential field method

Motion reconstruction provides essential inputs for analyzing human movement through musculoskeletal simulations. To reconstruct joint angles from motion capture data, several multibody kinematic optimization methods have been developed. However, a computationally efficient method yet simple to implement while ensuring consistent kinematics at all levels is lacking. Here, we propose a potential field method generated by virtual spring-dampers connecting measured-derived skin markers to segment-fixed model points to reconstruct motion in a forward dynamic manner by solving the equations of motion. The virtual spring-damper forces move the mechanical system to minimize the elastic potential and the distance between markers during the motion. Several evaluation strategies are performed which demonstrate that the potential field method is computationally fast (2.5ms per frame) with comparable accuracy to the well-established least squares method in terms of reconstructed marker trajectories and joint angles (RMSE < 0.37 mm, 1.87°) and with low marker residuals (< 18.7 ± 12.6 mm) in line with reported ranges. Furthermore, soft tissue artifacts are compensated well compared to the simulated true values (RMSE < 1.66 mm, 3.69°). Sternoclavicular, scapulothoracic and glenohumeral rotations were reconstructed well the major trends and magnitudes of experimental measurements. We anticipate our method will pave the way for complex applications that demand reliable and rapid large-scale biomechanical analysis of human movement.