Support Vector Machines for Inferring Distracted Behavior of Drivers Wearing Smart Glasses

摘要

Driver distraction refers to the lack of attention to the driving tasks due to engagement in secondary tasks. Most methods reported in the literature are based on visual-features analysis of head pose, since it is a strong indication of driver distraction. In contrast, we propose to use the inertial sensors embedded in smart glasses. To this aim, we collected data from five participants and conducted experiments to assess the feasibility of using support vector machines (SVM) to generate drivers' models to infer their focus of attention. The results show that using the personalized training model renders an acceptable accuracy to identify particular car cabin's spots where drivers focus their attention (i.e. accuracy was greater than 50 % and less than 81.44 % for all subjects).