Hypo-Driver: A Multiview Driver Fatigue and Distraction Level Detection System

Traffic accidents are caused by driver fatigue or distraction in many cases. To prevent accidents, several low-cost hypovigilance (hypo-V) systems were developed the past based on a multimodal-hybrid (physiological and behavioral) feature set. Similarly this paper, real-time inattention (Hypo-Driver) detection system is proposed through multi-view cameras biosignal sensors to extract hybrid features. The considered features derived from non-intrusive that related changes driving behavior visual facial expressions. get enhanced uncontrolled environment, three deployed multiview points (0°, 45°, 90°) of drivers. develop Hypo-Driver system, physiological signals (electroencephalography (EEG), electrocardiography (ECG), electro-myography (sEMG), electrooculography (EOG)) behavioral information (PERCLOS70-80-90%, mouth aspect ratio (MAR), eye (EAR), blinking frequency (BF), head-titled (HT-R)) collected pre-processed, then followed selection fusion techniques. behaviors classified into five stages such as normal, fatigue, inattention, cognitive drowsy. This improved hypo-Driver utilized trained convolutional neural network (CNNs), recurrent long short-term memory (RNN-LSTM) model used After these features, hypo-V layers dropout-layer deep-residual (DRNN) model. test performance data 20 drivers acquired. results compared state-of-the-art methods presented. Compared Hypo-V average, achieved accuracy (AC) 96.5%. obtained indicate multimodal outperforms other handling anomalies.