A Motion-Capture-Based Method for Upper-Limb Joint-Angle Estimation

Abstract Accurate and efficient upper-limb joint-angle estimation is important for human motion analysis and humanoid robot control. A rigid-body-pose-based method is presented for estimating the seven rotational degrees of freedom of the human arm using a multi-camera optical motion-capture system. Three non-collinear markers on each segment are used to reconstruct the orientations of the torso, upper arm, forearm, and hand. A seven-degree-of-freedom D-H model is established, and shoulder, elbow, and wrist angles are analytically solved from the reconstructed segment orientations. Constraint-based solution selection resolves multiple inverse-kinematic solutions, and Kalman filtering suppresses noise and motion fluctuations. Comparative simulations show that, under 2.0 mm marker noise, the proposed method achieves a mean RMSE of 2.049°, comparable to numerical inverse kinematics at 2.048° and lower than Euler-angle decomposition at 5.228°. Its computation time is only 0.0101ms per frame, representing a substantial improvement in computational efficiency compared with numerical inverse kinematics. Vicon experiments further confirm the feasibility of continuous seven-degree-of-freedom joint-angle estimation.