Head-Free, Remote Eye-Gaze Detection System with Easy Calibration Using Stereo-Calibrated Two Video Cameras

Abstract

The video-based, head-free, remote eye-gaze detection system based on detection of the pupil and the corneal reflection was developed using stereocalibrated two cameras. The gaze detection theory assumed the linear relationship; θ = k|r′|. Here, θ is the angle between the line of sight and the line connecting between the pupil and the camera, and |r’| indicates the size of the corneal reflection - pupil center vector. Three novel easy calibration methods were proposed; ‘automatic’, ‘one-point’, and ‘two-point’. In the ‘automatic’, the user does not have to fixate the specific location in the PC screen. In the ‘one-point’, the angular difference between the optical and visual axes of the eye was determined and used for compensation. The ‘two-point’ was proposed to compensate the nonlinear relationship between |r’| and θ, which occurs when θ is large. The precision of gaze detection was compared among the three methods using the developed gaze detection system.