The goal of this project is to create a single-camera vision system to monitor the behavior of automobile drivers. The system is to distinguish between safe and unsafe driving behaviors. Examples of unsafe driving behavior include talking on a cellular telephone, adjusting the controls of the in-car stereo, and eating, all while driving. The motions and positions of the arms and hands while performing these actions are very different than they are for regular safe driving. The method uses these cues to automatically distinguish between different driver behaviors, and output the time intervals during which the driver performed these different behaviors.

Human skin-tone falls into a certain region of color space regardless of race. Using this knowledge, a skin tone mask can be extracted for each frame. These masks are then clustered based on low-motion periods in the video. A Bayesian learning method is then used on the clusters to distinguish between safe and unsafe driver behavior.

Publications

• Harini Veeraraghavan, Nathaniel Bird, Stefan Atev, and Nikolaos Papanikolopoulos, "Classifiers for driver activity monitoring," Transportation Research: Part C, vol. 15, no. 1, pp. 51-67, February 2007.
• Harini Veeraraghavan, Stefan Atev, Nathaniel Bird, Paul Schrater, and Nikolaos Papanikolopoulos, "Driver activity monitoring through supervised and unsupervised learning," IEEE International Conference on Intelligent Transportation Systems, pp. 580-585, September 2005.