Designing an empirical framework to estimate the driver's attention

Abstract

Driver inattention is thought to cause many automobile crashes. Therefore, it is really important to pay high attention all the time though, for one split second of distraction can result in a terrible accident. However, developing an automatic driver assistance system for monitoring his/her attention during driving is a challenging task in computer vision. Level of attention of the driver may vary from high to low in different situations. Serious accident may be happened while the driver's attention becomes very low in certain cases. In this paper, we propose a vision based framework to estimate the level of attention of driver during driving. We have used Kinect sensor to collect the face angle and lip motions information to estimate level of attention. Moreover, the system generates an alarm as awareness signal of the driver in case of low attention level. Onboard experiments demonstrate the effectiveness of the approach, even in case of adverse light conditions. Experiments with 10 driver's show that the system is functioning quite well and measure the driver's attention level with reasonable accuracy.