Abstract
In this thesis, we continue the research of [Kuo, 1997] to complete a human-computer interface for disabled persons. We develop an algorithm to extract four feature points (two pupils and two inner eye corners) from the visible-light gray-level CCD image of the user's face and to estimate the position at which the user is gazing on the monitor. The advantage of this algorithm is that it takes only 0.2 seconds on a normal personal computer (Pentium - 166) to grab an image from CCD camera, extract feature points from the image and to estimate the position at which a user is gazing at. We develop a fast-response system with 3*3 icons on the monitor. After the training stage, the user can gaze at one of the nine icons to make the selection as if he is using a mouse. We also show three application examples to demonstrate the possibility of this technique in helping the disabled persons in communication, operating tools, and entertainment.