Publication
Intelligent Sensor Fusion with Online Distributed MIMU Calibration for Wearable Motion Capture
Hammad Tanveer Butt; Mathias Musahl; Manthan Pancholi; Pramod Murthy; Maria Alejandra Sanchez Marin; Didier Stricker
In: 22nd International Conference on Information Fusion (Fusion-2019), IEEE. International Conference on Information Fusion (FUSION-2019), July 2-5, Ottawa, Ontario, Canada, IEEE, 7/2019.
Abstract
Wearable systems for human motion capture based on low cost micro-electromechanical magnetic-inertial measurement units (MIMUs) show degraded performance when these systems are run for long time, especially indoors. The limitation arises due to the inhomogeneous magnetic field in the indoor environment, the high dynamic accelerations of body segments and due to bias variation of low cost sensors over time and with temperature. In our approach, we perform sensor fusion and online self-recalibration of accelerometers, magnetometers and rate gyros in a distributed algorithm, in order to provide an accurate orientation estimate. We developed a novel approach to use quasi-static time steps for calibration of rate gyros. To do this, an auto-calibration algorithm runs on each sensing node and provides correction to rate gyro biases. Together a hybrid sensor fusion and calibration algorithm is run on the hub for every sensor node. Our sensor fusion uses a novel adaptive covariance based EKF which makes it robust to both dynamic body accelerations as well as disturbed magnetic field. It also intelligently updates the accelerometer and magnetometer, when disturbances are low. This self-recalibration feature of our algorithm helps to achieve precise orientation estimates in highly dynamic conditions and avoid drift or error accumulation in inhomogeneous magnetic fields.