Eye Detection for near Infrared Based Gaze Tracking System

Abstract

This paper presents eye detection method for gaze tracking system using the features of eye and corneal reflection. The proposed method sequentially discards the regions that are expected not to be an eye by classifiers based on the features such as pupil intensity and appearance, corneal reflection intensity, and so on. The classifiers are designed with some empirical parameters, thus the proposed method does not need training process.