Depth of Eye Gaze Estimation using 3D Eye Tracking using Deep Learning

Document Type : Original Article

Authors

1 University of Tabriz

2 university of Tabriz

Abstract

Recently, with the growth of artificial intelligence technologies, research in the field of HCI has grown increasingly. Eye gaze tracking using machine vision and image processing has attracted the attention of researchers due to its advantages such as non-intrusive and no use of harmful radio and magnetic signals, cheapness, and portability. In this article, a new method for estimating the observer's gaze depth using 3D eye tracking with linear regression and deep learning methods is presented. A Valuable data set was prepared for the first time, and in it, a large number of images of the eyes of people, of various ages and genders in variable lighting conditions looking at a moving target at different distances, were recorded. A deep learning-based neural network was used to increase the accuracy of eye feature extraction. The idea of tracking the physiological changes of the eye, including the location and size of the pupil the distance between the pupils of the two eyes, and the change in the location of the reflective points of the cornea due to changes in the brightness of the environment and the distance of the target from the eye, also phenomena such as perspective difference and adaptation, was used to detect the observer's gaze depth. The results showed that the depth of the observer's gaze is estimated with the linear regression method with an accuracy of over 90% and with the deep learning method with an accuracy of nearly 93%.

Keywords

Main Subjects

Advanced Signal Processing

Articles in Press, Accepted Manuscript
Available Online from 29 August 2024

Files

Share

How to cite

Statistics