Direction estimation model for gaze controlled systems

Abstract

Gaze detection requires estimation of the position and the relation between user's pupil and glint. In this research paper, a Gaze Direction Estimation (GDE) model, a feature based shape method has been proposed for the comparative analysis of two standard edge detec-tors canny and sobel for estimating position of the glint coordinates and subsequently gaze direction based on the different human eye images dataset. The results indicate fairly good percentage of the cases where the correct glint coordinates and correct gaze direction quadrants have been estimated by the canny edge detector as it performs better than the sobel operator in most cases. These results can further be used for improving the accuracy and performance of different eye gaze based controlled systems.