Inertial-based real-time human upper limb tracking using twists and exponential maps in free-living environments

This paper presents a novel method for real-time human upper limb tracking with MARG sensor arrays in free-living environments. An adaptive-gain orientation filter as our previous work, was applied to a low-cost MARG sensor array. Based on an efficient Gauss-Newton optimization algorithm (GNA), particularly for reducing the computing load using only one iteration, this paper firstly proposes a novel method which is analytically derived for robust real-time orientation estimation to not only calculate the error direction of the gyroscope measurements but also handle magnetic distortions. Secondly, a human upper limb motion tracking system applying twists and exponential maps is well-developed based on the preestimated orientation, where a human kinematic model for upper extremity is designed for conveniently estimating joints' position and orientation within global frame. The performance of the proposed approach is verified by simulation and experimental results compared with the Vicon motion capture systems, which validate its effectiveness and accuracy.