Online Calibration Procedure for Motion Tracking with Wearable Sensors Using Kalman Filtering

Abstract

Several methods have been developed to estimate human limbs poses through inertial measurement units (IMUs). Although a big effort has been dedicated to the selection of the sensor fusion algorithm, less attention has been paid to some aspects that may cause considerable estimation errors. This paper presents a novel method for human upper limb motion tracking that accounts for the motion of the IMUs with respect to the attached limb and a calibration procedure for the estimation of limbs lengths. Three Unscented Kalman Filters are proposed to estimate joint angles, IMUs poses, and the limbs lengths based on the IMUs measurements. We validate our method by means of an optical motion tracking system that we used to calculate wrists position. This approach shows to be able to estimate unknown link lengths, to update correctly IMUs position and to improve the wrists position estimation.