Polarization has been considered to be a reference instead of near-infrared light sources in eye tracking because the light emitted from a liquid-crystal display is typically polarized. However, the degree of polarization depends on the display content. Thus, devising a novel method is crucial for stably extracting the display reflection from the corneal surface. Therefore, we propose an eye-tracking method that inserts a white background between the display contents using a high-speed display to extract the screen reflection on the cornea. A high-speed camera and polarization modulator are integrated to extract the polarized light emitted from a high-speed display, and then the point-of-gaze is estimated. We evaluated the accuracy of the estimated point-of-gaze under several conditions to compare the proposed method with conventional approaches. The results revealed that the proposed method improved eye gaze estimation.
Bingjie Xu Suzhou Industrial Park Institute of Vocational Technology, YijiaAn Xi’an Jiaotong-Liverpool University, Qinglei Bu Xi’an Jiaotong-Liverpool University, Jie Sun Xi’an Jiaotong-Liverpool University
Hannah Friederike Fischer German Research Center for Artificial Intelligence, Anke Königschulte German Research Center for Artificial Intelligence, Jana Koch C&S Computer and Software, Serge Autexier German Research Center for Artificial Intelligence, Gesche Joost German Research Center for Artificial Intelligence
Yaxuan Liu National University of Singapore (Suzhou) Research Institute, YijiaAn Xi’an Jiaotong-Liverpool University, Keming Zhang National University of Singapore (Suzhou) Research Institute, Martijn ten Bhömer Xi’an Jiaotong-Liverpool University, Qinglei Bu Xi’an Jiaotong-Liverpool University, Jie Sun Xi’an Jiaotong-Liverpool University, Siyuan Chen National University of Singapore (Suzhou) Research Institute