Fusion of eye-tracking data from multiple observers for increased 3D gaze tracking precision

Abstract

In this paper we discuss an approach to extract 3D gaze data information from binocular eye-tracking data. Factors such as tracking noise, tracking precision and observation distance limit the resolution of gaze tracking in three dimensions. We have developed a methodology, which uses a model of the stereoscopic human visual system (HVS) to analyze per-eye gaze data and to convert it into a something we call stereoscopic volume-of-interest (SVOI). We have found that using data from multiple observers increases the tracking precision. We aim to find the link between number of observers and tracking precision. This would allow one to optimize the number of participants involved in 3D gaze-tracking experiment, in order to achieve certain level of 3D gaze tracking precision.