Deep System For Physio-To-Sobriety Augmented Driving Risk Assessment In Next Generation Cars

Recent advances in automotive field confirmed that the major car makers deploy significant effort on the assessment of car driving safety. In this field, the authors propose an innovative approach that combine a physio-based drowsiness assessment of the car driver with an adaptive sobriety estimation through an intelligent electronic sensing system. Specifically, a coupled physio-probe embedding Near infra-Red (NiR) LEDs with a Silicon Photo-Multiplier (SiPM) detector is proposed. The implemented physio-probes hosted in the car steering, allow to detect a back-scattered PhotoPlethysmoGraphy (PPG) bio-signal of the driver to be used to retrieve a robust drowsiness estimation. In parallel, a further deep network will be able to learn specific embedded alcohol-dynamic features extracted from the car driver breath sampled by ad-hoc enhanced air-quality sensor. The experimental results (overall accuracy of 98 %) confirmed the effectiveness of the proposed system.