Development of technologies to monitor the state of the driver is essential in order to provide appropriate services for various driving situations. For the last decade a variety of driver state monitoring techniques have been proposed from many studies. Driver state monitoring systems generally work based on driving patterns, driver’s video or physiological signals. Driver’s video or driving patterns are convenient to acquire, but to assess driver state accurately is difficult because these methods assess the driver state indirectly. On the other hand, the analysis based on driver's physiological signals can monitor the state of the driver directly, but the sensors are not adopted due to the sensor's low usability in vehicle environment.
The proposed driver state monitoring system aims to assess driver’s drowsiness, fatigue, and distraction accurately while achieving high usability through analyzing the driving patterns and video of the driver together. The driver state monitoring system based on driving patterns is able to see the trend of driver state, but it is difficult to determine exactly when the driver is in a dangerous situation, like a microsleep. On the other hand, the video based driver state monitoring system makes it easy to determine the moment of falling asleep, but it needs an additional logic limiting the detection range to prevent increasing a wrong detection rate. The proposed logic finds drowsy driving sections by analyzing the driving patterns, and determines exact time when the alarm is triggered by analyzing the driver’s video. This configuration makes the proposed logic decide driver state with a high accuracy and provide an alarm within an appropriate time. This study is preliminary to validate a possibility of the proposed algorithm. The proposed driving pattern based algorithm was validated by comparing with the self-assessment and driver’s physiological signal. And the facial image based algorithm achieves a high accuracy of detecting face direction and eye blink.