Physiological signal based detection of driver hypovigilance using higher order spectra

Abstract

In recent years, driver hypovigilance which includes driver drowsiness and driver inattention is one of the major causes of road accidents and can lead to severe physical injuries, deaths and significant economic losses. Reliable driver hypovigilance detection system which could alert the driver before a mishap happens would ensure less road accidents. Previous research works have reported only on detecting either drowsiness or inattention. In this work, the focus is on developing a system that can detect hypovigilance, which includes both drowsiness and inattention, using Electrocardiogram (ECG) and Electromyogram (EMG) signals. Researchers have attempted to determine driver drowsiness or driver inattention using the following measures: (1) subjective measures, (2) vehicle-based measures, (3) behavioral measures and (4) physiological measures. A detailed review on these measures as to the sensors used, advantages and limitations associated with each measure is provided. The different ways in which drowsiness and inattention has been experimentally manipulated is also discussed. ECG and EMG signals are less intrusive as compared to other physiological signals and provide true state of the driver. Drowsiness has been manipulated by allowing the driver to drive monotonously at a limited speed for long hours and inattention was manipulated by asking the driver to respond to phone calls and short messaging services. A total of 15 male subjects participated in the data collection process and drove for two hours in a simulated environment, at three different times of the day (00:00 – 02:00 hours, 03:00 – 05:00 hours and 15:00 – 17:00 hours) when their circadian rhythm is low. ECG and EMG signals along with the video recording have been collected throughout the experiment. The gathered physiological signals were preprocessed to remove noise and artifacts. The hypovigilance features were extracted from the preprocessed signals using conventional statistical, higher order statistical and higher order spectral features. Statistically significant differences were observed between the alert, drowsy and inattentive states in both the physiological signals. The features were classified using k nearest neighbor, linear discriminant analysis and quadratic discriminant analysis. The energy feature of ECG signals gave a maximum accuracy of 93.35 %. The bispectral features gave an overall maximum accuracy of 96.75 % and 92.31 % for ECG and EMG signals respectively using k fold validation. The features of ECG and EMG signals were fused using principal component analysis to obtain the optimally combined features and the classification accuracy was 96%. In case of drowsiness, the driver has to be alerted on time. Hence, the different stages of drowsiness were classified with an overall accuracy of 71 %. Alerting the driver during initial stage of drowsiness would minimize accidents. In the future, the performance of hypovigilance detection system can be enhanced my merging these physiological measures with behavioral measures and vehicle based measures. A hybrid drowsiness detection system that combines non-intrusive physiological measures with other measures would accurately determine the drowsiness level of a driver. A number of road accidents can be avoided if an alert is sent to a driver who is drowsy or inattentive.