Dynamic accuracy assessment of data-fusion techniques for wearable, inertial and magnetic based human motion capture

Abstract

Wearable motion capture systems, based on magneto and inertial sensors are considered a promising technology in the field of human motion capture. Compared to golden standard systems, e.g. based on stereo-photogrammetry, they feature a better portability and usability and are suitable for `out of the lab' data collection. While different applications of human motion capture systems based on magneto and inertial sensors have been explored in literature, few studies have provided insights about the accuracy that can be achieved using data-fusion techniques and even less did that in a rigorous and controlled way. This work focuses on the dynamic absolute accuracy assessment of data-fusion algorithms for orientation tracking under controlled conditions. A 7 DOF robotic arm is used in order to apply controlled angular displacements to the motion capture system, in a range of frequencies that is of interest to human motion capture applications. Preliminary results on absolute accuracy that can be achieved with popular and open-source available fusion algorithms, running on a commercial hardware, are reported.