A Study on Reducing the Detection Errors of the Center of the Iris for an Eye-Gaze Interface System

Abstract

We have developed an eye-gaze interface system. The purpose of this system is to develop an easy to use system for spreading to ordinary people as well as people with motor-disabilities of upper limbs. Our system uses a compact video camera and a Windows PC equipped with a frame grabber.
This system detects the eye-gaze position on a computer display through the detection of the center of the iris from a captured eye image. When a user gazes at a peripheral point on the display, eye-gaze detection has slightly poorer accuracy because of influences of inner and outer corners of the eye. In this paper, we propose two new procedures to overcome the problem. We show the evaluation of these procedures through the comparing experiments on the accuracy of the eye-gaze estimation.