A Straightforward IMU-Based Upper Limb Inverse Kinematics Compared with Open-Source Vision Algorithms

This study presents and validates an effective approach for computing upper limb joint angles from inertial measurement units (IMUs), using a custom inverse kinematics (IK) algorithm. Two commercial IMU systems, Xsens and Muse 221e, were employed and compared in terms of joint angle reconstruction accuracy. In recent years, wearable IMUs have emerged as a promising alternative to marker-based motion capture (MOCAP) and vision-based pose estimation methods, offering low cost, portability and fewer privacy concerns. However, sensor fusion challenges, such as reliance on magnetometer or alignment issues, can affect measurement reliability. In the proposed method, IMUs were directly aligned with the OpenSim model axes to minimize computational effort, and the resulting quaternions were processed via a custom IK algorithm that accounts for anatomical constraints and joint topology. Comparative analyses were conducted between the two IMU systems and with a camera-based AI solution, Sports2D. Muse, which does not rely on a magnetometer, showed higher errors. Across all comparisons, elbow joint estimates were consistently more accurate than those of the shoulder, confirming the latter’s greater biomechanical complexity. While the computed errors are higher than those from MOCAP, the method proved effective in capturing motion patterns and time-based features, highlighting the potential of low-cost IMUs for practical joint kinematics monitoring.