Identifying Distraction: Kinematic Detection of Off-Road Eye Glances

Visual attention is a key component of the driving task. Identifying times when the driver is not looking toward the forward views has been shown to be a successful indicator of distracted driving (Olson, Hanowski, Hickman, & Bocanegra, 2009). Previous technologies have successfully captured eye-glance location, but have come at great expense in terms of both cost and intrusiveness. This project explored the possibility of using a low cost and unobtrusive Inertial Measurement Unit (IMU). Through data collected from the IMU, the aim was to identify vehicle kinematic signatures associated with times the driver is looking away from the forward view. Two methods were utilized. The first of the two methods relied on yaw rate anomalies to detect steering corrections. The second utilized the detection of sharp lateral accelerations which occurred as a result of corrective steering maneuvers. These methods both showed promise, as the yaw rate anomalies method and the lateral acceleration anomalies method both showed improved detection rates over the baseline.