Abstract
Since driver's drowsiness leads to serious accidents, it is imperative to prevent drowsy driving. Now, the event data recorders are equipped in most of taxis and buses for imaging inside and outside the vehicles. We therefore propose a detection method of driver's drowsiness by using time-series data of event data recorder for applying the active safety system.In the experiment, 4 subjects drove on a driving simulator until they fell asleep. The driver's drowsiness level was evaluated based on facial expression every 10 seconds. Analyzing drowsiness level and logged data of an event data recorder, we extracted features from forward road images. The features were highly related to increasing drowsiness and applicable to drowsiness detection in real time. Also, we included time-series variation of data obtained from the event data recorder considering the fact that the increasing drowsiness is temporally continuity.
