Predicting Future Eye Gaze Using Inertial Sensors

Eye tracking is a technology that in high demand, especially for next-generation virtual reality (VR), because it enables foveated rendering, which significantly reduces computational costs by rendering only the area at user gazing resolution and rest lower resolution. However, conventional eye-tracking technique requires per-eye camera hardware attached near eyes within VR headset. Moreover, detected eye gaze follows actual with finite delay of latency, need image processing, system’s native latency. This paper proposes an solution predicts user’s future using inertial sensors are already built into headsets head tracking. To this end, we formulate three time-series regression problems to predict (1) current past orientation data, (2) (3) data only. We solve first second machine learning models develop two solutions final problem: two-stage single-stage approaches. The approach problem relies on connected series, one other problem. uses single model directly from data. evaluate proposed based real traces captured headset multiple test players, considering various combinations models. experimental results show reduce error center-fixed up 50% 20% anticipation times 50 150 ms, respectively.