Bright pupil-based pupil center tracking using a quadrant photodetector

Abstract

Pupil trackers typically use cameras mounted on head-worn devices that operate at 30-200 fps for near-eye devices and consume significant power and compute resources using various AI and image processing algorithms. Accurate and high-speed pupil tracking without a camera is challenging because of ambient light and pupil size variations. For gaze tracking, finding the location of the pupil center is sufficient. In this paper, we developed a quadrant photodetector (QPD) based pupil gaze tracker, which is simple, low-power, and fast. Our QPD approach is benchmarked with camera-based measurements. We calculated the optimal QPD size and the maximum detectable range as a function of pupil size using data from different subjects. The pupil center can be found with < 0.5 mm accuracy with a range of 7 mm and 10 mm for pupil sizes of 3.5 mm and 7 mm, respectively. Compared to a camera with > 10,000 dedicated pixels for imaging a pupil, the QPD approach uses only 4 photodetectors. The approach produces accurate and fast results at low power with negligible computational load. However, the QPD approach is limited in field-of-view and accuracy, which can be improved by increasing the number of detectors from 4 to 16 or more, which will still be orders of magnitude more efficient than using a camera.