Local quaternion weighted difference functions for orientation calibration on electromagnetic trackers

The accuracy of the electromagnetic tracking systems has been always an important issue with application to motion and kinematic analysis. Applications in virtual reality and gesture recognition require not only of improved accuracy but also fast error compensation. Several analytic methods have been used in order to correct the position error and they are well known and fast: polynomial fitting, calibration tables, and more recent, neural networks. We are interested in the orientation calibration of working spaces with possible high distortion conditions. Such conditions are prevalent in virtual environment spaces such as the CAVE and it is not always possible to avoid metallic components in the surroundings. In this paper, we introduce a calibration method for a multiple-sensor electromagnetic tracking system in an environment with highly electromagnetic distortional conditions. The target system is a twelve-sensor Ultratrak Polhemus Inc.^trade system. We compare two possible formulations: global parameter estimation and local parameter estimation for the corrective functions. It is assumed that the inverse quaternion error Q^-1isin exists and it is a function of the three-dimensional location: Q^-1 isin rarr f(x,y,z)