Analysis of Short Term Path Prediction of Human Locomotion for Augmented and Virtual Reality Applications

Abstract

When human locomotion is used to interact with virtual or augmented environments, the system's immersion could be improved by providing reliable information about the user's walking intention. Such a prediction can be derived from tracking data to determine the future walking direction. This paper analyses how tracking data relates to navigation decisions from an egocentric view in order to achieve a reliable and stable path prediction. Since tracking data is noisy, a smoothening is required that eliminates oscillations while still recognizing trends in human locomotion. Thus, we analyze different approaches for path prediction, determine relevant setting values, and verify the results by a user study. Results indicate that robust short term prediction of human locomotion is possible but care must be taken when designing such a predictor.