Dynamically consistent inverse kinematics framework using optimizations for human motion analysis

Human motion analysis is of crucial importance in countless applications such as human-robot interaction or during the design of assistive devices. Human motion should be estimated as accurately as possible at both kinematics and dynamics levels. Accurately estimating joint trajectories, inter-segmental loads, geometric parameters and body segment inertial parameters specific to each subject will allow most of the indexes used to quantify/analyze human motion to be reconstructed. In this study, a multi-objective optimization problem is formulated to estimate the joint angles, velocities, and accelerations. Moreover two constraint quadratic problems are used to determine geometric parameters and body segment inertial parameters. Contrary to state of the art methods that rely solely on marker data to perform inverse kinematics, the proposed approach relies on force-plate data to obtain dynamically consistent joint trajectories. The proposed approach is evaluated on a squat exercise, performed by 8 subjects, and shows an improved accuracy in joint kinematics and inertial parameter estimation over the classical methods.