Real-Time Distraction Detection Based on Driver’s Visual Features

Abstract

Driver’s distraction has been listed as the leading contributing factor to traffic accidents for the past decades. This paper focuses on developing an approach to detect distraction real time by analyzing driver’s visual feature from the face region. The proposed approach uses visual features such as movement of eye and head to extract critical information to detect driver attention states and to classify it as either attentive or distracted. Deviation of eye center and head from their standard position for a period of time is considered to be useful cues for detecting lack of attention in this approach. At first face detection is performed after which region of interest (ROI) - eye and head region, are extracted using facial landmarks and lastly, head and eye movements are detected to classify attention state. To evaluate the system performance, we conducted an experiment in a real driving environment with subjects having different characteristics. Our system achieved on average 92% accuracy in detecting attention state for all tested scenarios.