Driver Distraction Detection and Identity Recognition in Real-Time

Drivers attending to primary driving tasks show specific eye and head movement behaviours, while the distracted drive generally covers the states including drivers´ eyes off the road and long-term eye closure. This paper presents a distraction detection system by using the strategy of ``attention budget´´. The states of eyes off the road and face with closed eyes are used to lessen the ``attention budget´´ while the reversed conditions gain it. Drivers´ gaze estimation is derived from the head motion, and the stage classifiers working with haar-like features are used to detect head movements and eye states. With regard to the factors of drivers´ personal characteristics in distraction detection, the recognition of drivers is implemented by extraction and matching of scale invariant feature transform features in detected frontal face. The results of experiments validate the effectiveness and robustness of the system.