Real-time calibration-free autonomous eye tracker

Abstract

In several fields of medicine, transportation and security the instantaneous gaze direction of a person under supervision is of crucial importance. We developed a contactless stereoscopic video-based eye tracker which works without any individual calibration. In real time it delivers information about the gaze direction frame by frame. The introduced algorithms are designed for computing the gaze direction within image acquisition time that is only limited by the hardware setup. The cameras are integrated into front-end modules by means of FPGA (programmable logic) circuits for image processing. Computation of the gaze direction is based on the spatial position of the pupil which is detected by a five-dimensional Hough transform. The system works under ambient light conditions whereas additional infrared illumination can be used to become independent of ambient light.