The identification of individualized eye tracking metrics in vr using data driven iterative- adaptive algorithm

Abstract

Eye tracking metrics provide information about cognitive function and basic oculomotor characteristics. There have been many studies analyzing eye tracking signals using different algorithms. However, these algorithms generally are based on the initial setting parameter. This might cause the subjective interpretation of eye tracking analysis. The main aim of this study was to develop a data-driven algorithm to detect fixations and saccades without any subjective settings. Five subjects were included in this study. Eye tracking signal was acquired with the VIVE Pro Eye in virtual reality (VR) environment while subjects were reading a paragraph. The algorithms based on the calculation of threshold were employed to calculate eye metrics including total fixation duration, total fixation number, total saccades number and average pupil diameter. The proposed algorithm, which is based on calculating the initial threshold, based on mean, and standard deviation of eye tracking signal within experiment duration, gave the same results obtained adaptive filtering reported in literature (average fixation duration for five subjects= 10634.6 ms ± 5117.9, average fixation count for five subjects= 21.8 ± 7.5). On the other hand, our proposed algorithm didn’t use any certain objective parameter as like adaptive filtering. As a conclusion, VIVE Pro Eye may be utilized as an eye movement assessment device, and, the suggested approach might be utilized to analyze objective eye tracking metrics.