A calibration process for tracking upper limb motion with inertial sensors

Real-time inertial tracking of human motion requires to attach inertial sensors to the major segments of a human body. Due to the intrinsic noises and drifts of the sensor data, as well as the non-rigidity of the human muscles and skins, the tracking could be unprecise, especially as time elapses. Therefore, an elaborate calibration process and a noise-insensitive tracking algorithm are needed. This paper presents an Extended Kalman Filter-based method for calibrating the relative orientation between human upper limb bones and the sensors attached to them. The calibration result, represented as a rotation matrix, is then used in the tracking of the arm motion with the knowledge of the physical segment kinematic model. Several experiments have been performed to validate the effectiveness of our method.