Mobile eye tracker

• Grant number: 13/06791-0
• Support Opportunities: Scholarships in Brazil - Doctorate (Direct)
• Start date: June 01, 2013
• End date: June 30, 2017
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Carlos Hitoshi Morimoto
• Grantee: Andrew Toshiaki Nakayama Kurauchi
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated scholarship(s): 14/12048-0 - Wearable eye gaze tracking, BE.EP.DD

Abstract

Eye Gaze Trackers (EGT) are devices that estimate the user's gaze position at each instant. The most common EGTs use video cameras, allowing for remote gaze observation, i.e., without any device in direct contact with the user. The introduction of micro-cameras facilitated the development of mobile head mounted EGT using a second camera for capturing scene images. The scene images are used to project the gaze information. Though there were recent developments there are still challenges that hinder the use of such devices. We can cite the high cost, lack of accuracy, the difficulty of calibration, and the need for controlled lighting conditions. This project aims to convert the master's scholarship into a direct doctorate scholarship. The focus of the master's project is to develop a low cost head-mounted mobile EGT robust to lighting condition changes. For the direct doctorate the project will be extended to research yet other central problems to the advance of mobile EGTs. In particular the calibration problems and parallax correction. The parallax caused by the distance between the scene camera and the eye has a direct effect on gaze estimation accuracy. Among the possible alternatives to be investigated we can cite the study of epipolar geometry to extract depth information using stereo cameras, or the use of invariant properties of the projective geometry that can compensate the paralax effects. The calibration of an eye gaze tracker allows the orientation of the eye to be mapped to objects in the scene. Traditional calibration methods depend on the collaboration of the user, who must look at a certain amount of points in the scene. A high amount of points is unwanted, as it complicates the calibration process. A too low amount of points might result in a less accurate mapping. In this project we will study techniques that minimize the number of points needed for calibration that still result in good accuracy, including techniques that allow for the calibration process to be hidden and that can be used to dynamically recalibrate the system to maintain its accuracy. The student, Andrew T. N. Kurauchi, has an excelent academic record. He has recently graduated from USP, where he obtained a B.Sc. in Computer Science. His final overall average grade was 8.66, and he scored maximum grade in 5 out of 6 disciplines he attended as a graduate student. In the first semester of his scholarship the student already developed an initial version of the eye gaze tracking software, which is already in use in one of the experiments of our group. In the end of the same period he also participated in the development of two international publications. The first is an extended abstract accepted in CHI 2013, and the second was accepted in a workshop organized by SIGEYE, to be held along with CHI 2013. (AU)

Academic Publications

(References retrieved automatically from State of São Paulo Research Institutions)

KURAUCHI, Andrew Toshiaki Nakayama. EyeSwipe: text entry using gaze paths. 2018. Doctoral Thesis - Universidade de São Paulo (USP). Instituto de Matemática e Estatística (IME/SBI) São Paulo.