Multisensor Integrated Prediction for Virtual Reality

Abstract

Unconstrained measurement of human head motion is essential for HMDs (headmounted displays) to be really interactive. Polhemus sensors developed for that purpose have deficiencies of critical latency and low sampling rates. Adding to this, a delay for rendering virtual scenes is inevitable. This paper proposes methods that compensate the latency and raises the effective sampling rate by integrating Polhemus and gyro sensors. The adoption of quaternion representation enables us to avoid singularity and the complicated boundary process of rotational motion. The ability of proposed methods under various rendering delays was evaluated in the respect of RMS error and our new correlational technique, which enables us to check the latency and fidelity of a magnetic tracker, and to assess the environment where the magnetic tracker is used. The real-time implementation of our simpler method on personal computers is also reported in detail.