Effective feature extraction from driving data for detection of danger awareness

Abstract

In recent years, the importance of driver’s support system is increasing as a solution for dealing with car related accidents. These driving support systems are equipped with functions for avoiding various hazards when the driver drives the vehicle, reducing the risk of causing an accident. In this research, we focus on the time series data of the driving behaviour of the driver, and based on these data, experiments aiming at development of the dangerous driving detection system due to cognitive distraction of the driver have been conducted. The driving behaviour data have been collected from driving simulator which contain driver’s actions mainly steering, accelerator and foot brake operations. It has been observed that the driving behaviour of each driver changes while driving in the state of distraction from while driving attentively and by analyzing these changes, the driver’s distraction from the normal state can be detected. The objective of this paper is to find the effective features for detection of distracted driving of specific driver in real time (specific short intervals). From the collected data of driving behaviour of multiple subjects, static feature based driving model and dynamic feature based driving model for individual drivers and all drivers for attentive driving and distracted driving have been developed. It can be shown from the results that distracted driving can be identified for individual in real time with stable accuracy using dynamic feature based models.