Abstract:
The use of human eye gaze as an input device for computer systems in place of a keyboard requires high resolution of screen positions. The previous vision-based eye gaze detection systems have not high enough resolution for such applications. This thesis proposes a method to determine the eye gaze positions on screen by using two-eye displacements as the information for mapping to a position on the screen. Images taken by a single camera positioned above the monitor while a user is looking at a target on the screen are used as the input of the proposed method. Firstly, the centers of the two eyes are determined. Then, the distances between the center of the eyes and the reference point on the user face are calculated. These distances are used in computing the displacement. Finally, given that the distance between the users face and the camera is known, the perspective projection is applied to map the displacement to a position on screen. A camera calibration is required for setting the distance between the users face and the camera. To evaluate the method, the experiments were performed on 20 persons. The distance between the users face and the camera was at 60 cm, and the screen was divided into 10x10 positions on a 17-inch monitor with the screen resolution of 1024 x 768 pixels. The result resolution was cm, which can be translated into to menu of 5x8 positions on a 17-inch monitor. Finally, the experiments were carried out with the menu of 5x8 and 7x10 positions on 4 of the 20 persons. The results showed 100 percent and 96 percent correction, respectively.