Abstract:
In this paper, we propose 'Hands-free Interface', a technology that is intended to replace conventional computer screen pointing devices for the use of the disabled. We describe a real time, non-intrusive, fast and affordable technique for facial feature tracking. Our technique ensures accuracy and is robust to feature occlusion as well as target scale variations and rotations. It is based on a novel variant of template matching techniques. A Kalman filter is also used for adaptive search window positioning and sizing. © 2004 IEEE.