Recent statistics confirmed that the car driver drowsiness monitoring reduces drastically the road accidents. In scientific literature, several advanced approaches have been proposed to monitor the driver's level of attention, providing a real-time warning to increase driving safety. With this aim, we propose an innovative method which consists of ad-hoc designed bio-sensing system to assess the car driver's physiological state. The designed bio-sensing system includes a probe which detects a physiological signal of the subject i.e. the PhotoPlethysmoGraphy (PPG). The physio-probe device has been embedded on several points of the car's steering wheel in order to sample the PPG signal from the driver's hand. Furthermore, ad-hoc motion magnification algorithm was developed to reconstruct PPG from visual car driver face motions when physical PPG signal is unavailable. An innovative deep learning system completes the proposed pipeline in order to classify the driver drowsiness from the so collected PPG signal. The drowsiness detection performance (average accuracy of around 90%) confirmed the effectiveness of the proposed approach.
Deep bio-sensing embedded system for a robust car-driving safety assessment
Spampinato C.;Conoci S.;
2020-01-01
Abstract
Recent statistics confirmed that the car driver drowsiness monitoring reduces drastically the road accidents. In scientific literature, several advanced approaches have been proposed to monitor the driver's level of attention, providing a real-time warning to increase driving safety. With this aim, we propose an innovative method which consists of ad-hoc designed bio-sensing system to assess the car driver's physiological state. The designed bio-sensing system includes a probe which detects a physiological signal of the subject i.e. the PhotoPlethysmoGraphy (PPG). The physio-probe device has been embedded on several points of the car's steering wheel in order to sample the PPG signal from the driver's hand. Furthermore, ad-hoc motion magnification algorithm was developed to reconstruct PPG from visual car driver face motions when physical PPG signal is unavailable. An innovative deep learning system completes the proposed pipeline in order to classify the driver drowsiness from the so collected PPG signal. The drowsiness detection performance (average accuracy of around 90%) confirmed the effectiveness of the proposed approach.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
