Illumination invariant head pose estimation using single camera

A learning based approach is presented for a single camera based robust head pose estimation in highly cluttered and complex real world environments. The method makes effective use of a recently introduced real-time 3D depth sensing technology, resulting in illumination-invariant head pose estimation. The main target application of our work is to classify the focus of attention of an automobile driver as straight-ahead, towards-rear-view mirror, towards the dash board, etc. This type of information is expected to be useful in conjunction with other sensors to enable safer driving.