Estimation of Inattention in Driving by Using Driver Head Pose and Vehicle Information

This work deals with the problem of detecting driver’s looking aside while driving. The main challenge is how to differentiate the two cases where drivers turn their faces as a part of normal driving behaviors, and where drivers get distracted for some reason and, as a result, take their eyes off a road. Our solution for this problem is to incorporate multiple observation cues in the dynamic Bayesian network framework such as the vehicle’s speed and relative orientation to a road lane in addition to a driver’s facial orientation. We explain the algorithm of our method and report experimental results conducted by using a driving simulator.