Vision Based Wearable Eye-Gaze Tracking System

This paper proposes a new wearable eye-gaze tracking system with a single webcam mounted on the glasses. First, the region of interest (ROI) of eye is extracted by skin detection and eyelid removing. Then Hough circle detection is used to search the candidate of circles in the eye's ROI and applied to determine the status of opening or closing of eye. Based on the circle detection, the eye center and radius are detected by using least square based starburst algorithm. To enhance the robustness of eye center estimation, we use particle filter to deal with the disturbance of eye center position induced by different lighting condition and noise. Furthermore, the gaze position on the screen is projected polynomial interpolation. The experimental results verified the accuracy and effectiveness of the proposed system.