Eye-gaze distance estimation based on gray-level intensity of image patch

Abstract

This paper proposes an eye-gaze distance estimation technique based on the gray-level intensity of a single image patch from a single camera. The proposed measuring system is easy to estimate the distance between a user's eye and a computer screen. For eye gaze distance estimation, the eye gaze image is reflected by two infrared light emitting diodes attached on the vertical center of the left and right frame of a computer screen. The images are captured using a camera with a lens with focal length of 25 mm. The proposed method is tested on five users. The results are 95.64 % accuracy using linear least squares method and 96.54 % accuracy using least squares second order polynomial. The absolute errors for distance estimation are 2.97 cm and 2.47 cm using the linear least square and the least squares polynomial, respectively. The comparison shows the proposed method performs better than the pixel counts method and the two-eye distance based method. © 2011 IEEE.