Head-mounted binocular gaze detection for selective visual recognition systems

For an effective and efficient visual recognition system, the region-of-interest extraction of users is one of most important image processing duties. Generally, the position of the pupil in an eyeball is directly related to the user's interest in an input visual image. However, when using a monocular eye tracking system, it is not easy to discern an accurate position of the region-of-interest of the user in a three-dimensional space. In this paper, an eye tracking system based on binocular gaze detection is presented. The proposed system is designed as a wearable device with three mini-cameras and two hot-mirrors for users to see through. The two eye monitoring system acquires the pupil images of both eyes through the hot-mirrors illuminated by infrared LEDs; the front-view camera acquires the image visible to the user. The experiment results show that the proposed system improves upon the region-of-interest localization error of the users in both the horizontal and vertical directions for the front-view images acquired by the user. Through the proposed system, it became possible to accurately extract the user's region-of-interest, and so can be used to improve the image processing speed through focused information processing for the region-of-interest and narrow the selective information acquisition needed for users.