A neural network based real-time gaze tracker
Journal of Network and Computer Applications
Hi-index | 0.00 |
A non-intrusive real-time program detects the face, theneyes and nose of a moving workstation user at a rate ofbetween 10 and 30 Hertz. Based on the head pose, it determineswhere on the display the subject is looking. The longterm goal is to provide a system for controlling a computerusing head movements and gaze direction. The current programcreates a base facility for other capabilities such asdetecting facial gestures, creating face models, and normalizingfor face recognition. A skin color model is usedalong with geometric knowledge about the face and weakassumptions about the lighting. Good results are reportedwith various subjects and conditions, including facial hair,3D motion, and use of eyeglasses.