Car-Driver Drowsiness Monitoring by Multi-layers Deep Learning Framework and Motion Analysis

Recent developments in the automotive industry have led to an interest in monitoring car driver drowsiness. The purpose is to develop an efficient system for the detection of bad psychophysical states in order to reduce the number of fatigue-related car accidents. Much of the current literature pays particular attention to the study of physiological signals to obtain information about cardiac activity by measuring the Heart Rate Variability (HRV). In fact, the HRV represents a useful indicator for evaluating physiological stress because it provides information about the cardiovascular system activity controlled by the Autonomic Nervous system. The present study is designed to analyze the skin micro-movements caused by blood pressure by extracting facial landmarks in order to reconstruct the photoplethysmogram (PPG) signal in a robust way. To conclude, we obtained evidence from the validation results to support the idea that the PPG signal detected by sensors and the reconstructed PPG by using facial landmarks are strongly correlated.