Estimating driver's readiness by understanding driving posture

In this paper, we propose a novel Driver Monitoring System (DMS) that estimates whether drivers are able to control vehicles. Provided with an on-board depth sensor and computing platform, the system keeps recognizing the driving posture. The system is trained beforehand to learn normal driving postures so that it can detect driver's inability state by calculating the “distance” between the known normal postures and the current posture. By using our relative representation, the system extracts the essential information of driving postures from depth map images. To increase the accuracy, each relative relationship among joints is weighted depending on its significance. We implemented our DMS on a real vehicle for evaluations. 10 participants drove the vehicle on three test courses and we checked our DMS performance in real time. The evaluation results show that the average false detection rate decreases to 0.01 %, so we confirmed that this novel DMS has potential performance for practical use.