A fast human face modeler and tracker for Driver Inattention Monitoring

For the application of Driver Inattention Monitoring System this paper propose a zero-order binocular EKF face tracker. By combining the facial feature extraction algorithm our proposed tracker could estimate the face pose, the eyeball yaw & pitch, the eyelids animation and the mouth & jaw animation as well as shape parameters. The facial feature extraction algorithms are also presented. Experiments shows the proposed full detection and tracking loop could give more reliable result than monocular tracker with real time performance on a general purpose personal computer. It is also shown the tracker has tolerance against feature point detection error.