| Summary: | Abstract In the last decade, significant advances have been made in sensing and communication technologies. Such progress led to a considerable growth in the development and use of intelligent transportation systems. Characterizing driving styles of drivers using in-vehicle sensor data is an interesting research problem and an essential real-world requirement for automotive industries. A good representation of driving features can be extremely valuable for anti-theft, auto insurance, autonomous driving, and many other application scenarios. This paper addresses the problem of driver identification using real driving datasets consisting of measurements taken from in-vehicle sensors. The paper investigates the minimum learning and classification times that are required to achieve a desired identification performance. Further, feature selection is carried out to extract the most relevant features for driver identification. Finally, in addition to driving pattern related features, driver related features (e.g., heart-rate) are shown to further improve the identification performance.
|
|---|
