In-attention state monitoring based on integrated analysis of driver’s headpose and external environment

Abstract

In Advanced Driving Assistance Systems (ADASs), for traffic safety, one of main application is to notify the driver regarding the important traffic information such as presence of a pedestrian or information regarding traffic signals. In a particular driving scenario, the amount of information related to the situation available to the driver can be judged by monitoring the internal infor‐ mation (for example driver’s gaze) and external information (for example infor‐ mation regarding forward traffic). Therefore, to provide sufficient information to the driver regarding a driving scenario it is essential to integrate the internal and external information which is lacking in the current ADASs. In this work, we employ 3D pose estimate algorithm (POSIT) for estimation of driver’s attention area. In order to estimate the distributions corresponding to the forward traffic we employ both bottom-up saliency map model and a top-down process using HOG pedestrian detection. The integration of internal and external information is done using the mutual information.