A classification method between novice and experienced drivers using eye tracking data and Gaussian process classifier

Abstract

We propose a classification method based on a binary Gaussian process classifier to classify novice and experienced drivers using eye gaze that can reflect drivers' attention and skill. Gaze behavior during lane changing task were collected from both novice drivers and experienced drivers by using an eye tracking system and a driving simulator in this study. We applied the Gaussian process classifier to the two-dimensional coordination data of the gaze behavior, and compared the performance of Gaussian process classifier with those of Gaussian mixture models that had the different number of components. Our proposed method showed the superiority in classification performance to the methods based on the Gaussian mixture models.