Sensor Fusion Based Real-time Measurement of Human Head Motion

Abstract

Most VR systems these days use unconstrained motion sensors like Polhemus Tracker, which has deficiencies of low sampling rate and critical latency. Some ad-hoc linear filters were proposed, but are known to be incomplete to rotatinal motion. We proposed the Sensor-Fusion-based method to the problem before. In this paper we improve this method by adopting quaternion and report its real-time implementation. We also propose to use normalized cross-correlation function to evaluate unconstrained sensors quantitatively.