We have presented a robust frame-rate pupil detector and eye gaze tracker, with high potential to be used in human-computer interaction, particularly in desktop and kiosk applications. The even and odd frames of a video camera are synchronized with two IR light sources. A pupil is alternately illuminated with an on-axis IR source when even frames are being captured, and with an off-axis IR source for odd frames. The on camera axis illumination generates a bright pupil, and the off axis illumination keeps the scene at about the same illumination, but the pupil remains dark. Detection follows from thresholding the difference between even and odd frames.
Once the pupil is detected, the corneal glint from the light sources is searched for, near the center of the pupil. The eye gaze tracker uses the center of the pupil and glint to estimate the position on the screen to where the user is fixating her/his gaze, after a brief calibration procedure that determines the mapping from coordinates of the pupil tracker to user screen coordinates.
The eye gaze tracker has been successfully tested for a very large number of people, and it has proven to be very robust. Future extensions include the generalization of the problem to a 3D model in order to allow for large head motion and enhancements on the pupil detector to increase its accuracy, that includes changes in the calibration procedure and mapping functions.