Beware the Sirens: Prototyping an Emergency Vehicle Detection System for Smart Cars

Abstract

Vehicle drivers should be able to react coherently in anomalous circumstances, such as the quick arrival of an emergency vehicle with sirens wailing. This situation requires all regular vehicles to give way or slow down, depending on the road and traffic conditions. In this paper, we address an automatic system that assists the driver in reacting to the arrival of an emergency vehicle by employing audio and video algorithms based on Deep Learning. More specifically, by leveraging sound recognition algorithms, the vehicle is able to detect the arrival of the emergency vehicle by its siren sound. In such an event, by making use of computer vision algorithms, the vehicle intelligence can monitor the driver’s gaze and awareness towards the emergency vehicle and assess his/her awareness. The paper describes the process of integrating these technologies into a commercial car, the creation of new datasets and the challenges encountered.