Hypovigilance detection by calculating and analyzing a spatio-temporal features of the face components based on 3D head orientation

The Drowsiness is the risk to fall asleep for one moment with the closed eyes, it is an intermediate state between the awake and the sleep. This state is involuntary and it is accompanied by a fall of vigilance. Consequently, the development of a system for the automatic control of driver tiredness, to prevent the driver in advance of the accidents, received an interest growing of research mediums. In this work, we propose an approache of drowsiness detection which makes it possible to determine the 3D head orientation to capture the lack of concentration states. This approach is based on the estimate of the head rotation angles in the three directions Y aw, Pitch and Roll by exploiting only three interest points of the face. The approache suggested are evaluated by the MiraclHB database. The evaluation show many promising results and show the effectiveness of the approache suggested.