Date of Award

Fall 2021

Publication Type

Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

Keywords

Eye-tracking, Reading, Eye gaze classification, Reading line classification algorithms

Supervisor

E. Kim

Supervisor

X. Chen

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract

In this thesis, we developed an algorithm to detect the correct line being read by participants. The comparisons of the reading line classification algorithms are demonstrated using eye-tracking data collected from a realistic reading experiment in front of a low-cost desktop-mounted eye-tracker. With the development of eye-tracking techniques, research begins to aim at trying to understand information from the eyes. However, state of the art in eye-tracking applications is affected by a large amount of measurement noise. Even the expensive eye-trackers still suffer significant noise. In addition, the inherent characteristics of gaze movement increase the difficulty of obtaining valuable information from gaze measurements. We first discussed an improved Kalman smoother called slip-Kalman smoother, designed to separate eye-gaze data corresponding to correct text lines and reduce measurement noise. Next, two different classifiers are applied to be trained; one is Gaussian discriminant based while the other is support vector machine based. As a result, our algorithm improved the performance of eye gaze classification in the reading scenario and beat the previous method.

Share

COinS