Face orientation estimation for driver monitoring with a single depth camera

Abstract

Careless driving is a major factor in most traffic accidents. In the last decade, research on estimation of face orientation and tracking of facial features through consequence images have shown very promising results of determining the level of driver's concentration. Image sources could be monochrome camera, stereo camera or depth camera. In this paper, we propose a novel approach of facial features and face orientation detection by using a single uncalibrated depth camera by switching IR depth pattern emitter. With this simple setup, we are able to obtain both depth image and infrared image in a continuously alternating grab mode. Infrared images are employed for facial features detection and tracking while depth information is used for face region detection and face orientation estimation. This system is not utilized only for driver monitoring system, but also other human interface system such as security and avatar systems.