Login | Register

Regression Based Gaze Estimation with Natural Head Movement


Regression Based Gaze Estimation with Natural Head Movement

Fu, Yang (2015) Regression Based Gaze Estimation with Natural Head Movement. Masters thesis, Concordia University.

[thumbnail of Final_submission_2015-05-28.pdf]
Text (application/pdf)
Final_submission_2015-05-28.pdf - Accepted Version


This thesis presents a non-contact, video-based gaze tracking system using novel eye detection and gaze estimation techniques. The objective of the work is to develop a real-time gaze tracking system that is capable of estimating the gaze accurately under natural head movement. The system contains both hardware and software components. The hardware of the system is responsible for illuminating the scene and capturing facial images for further computer analysis, while the software implements the core technique of gaze tracking which consists of two main modules, i.e., eye detection subsystem and gaze estimation subsystem.
The proposed gaze tracking technique uses image plane features, namely, the inter-pupil vector (IPV) and the image center-inter pupil center vector (IC-IPCV) to improve gaze estimation precision under natural head movement. A support vector regression (SVR) based estimation method using image plane features along with traditional pupil center-cornea reflection (PC-CR) vector is also proposed to estimate the gaze.
The designed gaze tracking system can work in real-time and achieve an overall estimation accuracy of 0.84º with still head and 2.26º under natural head movement. By using the SVR method for off-line processing, the estimation accuracy with head movement can be improved to 1.12º while providing a tolerance of 10cm×8cm×5cm head movement.

Divisions:Concordia University > Gina Cody School of Engineering and Computer Science > Electrical and Computer Engineering
Item Type:Thesis (Masters)
Authors:Fu, Yang
Institution:Concordia University
Degree Name:M.A. Sc.
Program:Electrical and Computer Engineering
Date:22 May 2015
ID Code:980073
Deposited By: YANG FU
Deposited On:02 Nov 2015 17:02
Last Modified:18 Jan 2018 17:50
All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Downloads per month over past year

Research related to the current document (at the CORE website)
- Research related to the current document (at the CORE website)
Back to top Back to top