An Evaluation of Extrapolation and Filtering Techniques in Head Tracking for Virtual Environments to Reduce Cybersickness

Currently, numerous users who employ HMD devices such as the Oculus Rift develop symptoms similar to motion sickness. Recent literature defines this phenomenon as cybersickness, and one of its main causes as latency. This contribution aims to analyze the accuracy of different extrapolation and filtering techniques to accurately predict head movements, reducing the impact of latency. For this purpose, 10 participants played aVRgame that required quick and subsequent head rotations, during which a total of 150.000 head positions were captured in the pitch and yaw rotation axes. These rotational movements were then extrapolated and filtered. Linear extrapolation seems to provide best results, with a prediction error of approximately 0.06 arc degrees. Filtering the extrapolated data further reduces the error to 0.04 arc degrees on average. In conclusion, until future VR systems can significantly reduce latency, extrapolating head movements seems to provide a low-cost solution with an acceptable prediction error, although extrapolating the roll axis movements remains to be challenging.