Iris tracking and blink detection for human-computer interaction using a low resolution webcam
Rupal Khilari · 2010
This paper proposes a technique to develop a non-intrusive interface for human-computer interaction based on visual cues of a human eye, captured using a low resolution USB camera. Such an interface can replace a traditional interface like a mouse thus helping the severely disabled to use the computer just as a normal individual would. It makes use of iris tracking and blink detection for this purpose. Considering fairly constant lighting conditions, a modified Hough Transform is used to track the iris at real time. The Between-The-Eyes feature serves as a stable reference point for gaze direction estimation. When mapped to the computer screen, incremental, directional movement of the mouse pointer position provides precise control over its movement. A method based on the projection function is proposed to detect the user's eye blinks, analyze their patterns and duration, interpreting them as appropriate mouse clicks. The system gives an overall accuracy of 87.4 %.