Noncontact video-based eye-gaze detection method allowing large head displacements

Abstract

The noncontact video-based eye-gaze detection method which allows large head displacements was developed. The eye was tracked by a two-dimensional camera rotation system. The distance between the camera rotation center and the eye was measured by an ultrasonic distance meter. A theory using trigonometry was proposed to determine the pupil center position in three-dimensional space. At this time, the displacements of the pupil from the camera frame center were considered to precisely determine the coordinates of the eye-gaze point on a computer display. The angle of eye's line of sight against the camera axis was determined from the relative positions of the centers of the feature points in the eye image, i.e., the glint (corneal reflection) and the pupil. The eye-gaze points while a subject fixated visual targets presented on the display were measured at video rate using custom-made image processors. Under the small head-movement condition and under the large head-displacements condition, the results showed almost no distortion between the target and eye-gaze positions. This indicated that the theory is adequate and the system worked well. The average error was approximately 0.7 deg in visual angle.