Detection of Driver Cognitive Distraction Using Machine Learning Methods

Driver distraction is one of the primary causes crashes. As a result, there great need to continuously observe driver state and provide appropriate interventions distracted drivers. Cognitive refers “look but not see” situations when drivers’ eyes are focused on forward roadway, their mind not. Typically, cognitive distractions can result from fatigue, conversation with co-passenger, listening radio, or other similarly loading secondary tasks that do necessarily take driver’s off roadway. This makes it hardest detect as no visible clues distraction. In this study, we have identified features different sources including eye-tracking, physiological, vehicle kinematics data relevant towards classification non-distracted drivers via analysis collected driving simulator study involving 40 across multiple scenarios. The key algorithms implemented include Random Forest, Decision Trees Support Vector Machines. A reduced feature set pupil area, vertical horizontal motion was found be predictive while maintaining an average accuracy 90% various road types. Additionally, impact types behaviour also identified. findings has practical application design monitoring systems.