Monocular head pose estimation using generalized adaptive view-based appearance model