A Detection Method of Driver's Face Orientation Based on Visual Cues and SVM

Face orientation can be used to derive vigilance cues about the vehicle driver. In this paper, a face pose estimation model based on visual features is presented. The face orientation is estimated through SVM with features detection such as eyes, mouth, facial outlines and so on. Innovative point of this research is the utilization of the features triangle, which composed with eyes and mouth locations, to obtain a robust face pose estimate. This method leads to the free face pose estimation and it is non-intrusive to the driver. Finally, the system is assessed with a commercial capture system. The evaluation results indicate that the system presented can be used in the driver assistant system.