Human-Computer Interaction Using Eye-Glasses Shape Information by 3-D Computer Vision Techniques

Abstract

An approach to human-computer interaction suing eye-glasses shape information by 3-D computer vision techniques is proposed. The direction of sight is estimated by the pose of a pair of eye glasses worn on the user. The eye-glasses shape used in this study is composed of two parallel lines and two circles. A monocular image of the eye-glasses shape is captured by a camera mounted on the computer monitor. The image processing and curve fitting techniques are employed to obtain the equations of lines and circles. The pose of the eye glasses with respect to the camera can be computed using the derived analytical solutions in terms of the coefficients of the lines and the circles. In addition, an approach to calibrating the geometric relation between the camera and the computer monitor is proposed. After knowing the pose of the eye glasses and the monitor plane equation in the image coordinate s stem, the intersection of the direction line of the eye glasses and the monitor can be used as input for some human-computer interaction applications. An advantage of this method is that the solution can be uniquely and an alytically determined without iterative computation. Experimental results of image processing and pose estimation show the effectiveness of the proposed approach.