AUTHOR=Arif Saad , Munawar Saba , Ali Hashim 

TITLE=Driving drowsiness detection using spectral signatures of EEG-based neurophysiology

JOURNAL=Frontiers in Physiology

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2023.1153268

DOI=10.3389/fphys.2023.1153268

ISSN=1664-042X

ABSTRACT=Drowsy driving is a significant factor instigating dire road crashes and casualties around the world. Its earlier and more effective detection can significantly reduce the lethal aftereffects and increase road safety. In this preface, A passive brain-computer interface (pBCI) scheme using electroencephalography (EEG) brain signals is developed for earlier detection of human drowsiness during driving tasks. This pBCI modality acquired electrophysiological patterns of 12 healthy subjects from the prefrontal cortex (PFC), frontal cortex (FC), and occipital cortex (OC) of the brain. Drowsiness activity was recorded using six EEG channels spread over the right and left hemispheres of the brain in PFC, FC, and OC. During the experiment, sleep-deprived subjects drove a vehicle in a driving simulator while their neurological state and physical behavior were continuously measured and recorded. In post hoc analysis, spectral signatures of the delta, theta, alpha, and beta rhythms were extracted in terms of spectral band power and band power ratios with a temporal correlation over the complete length of experiments. Minimum redundancy maximum relevance, Chi-square, and ReliefF feature ranking methods were employed and aggregated with a Ζ-score based approach to obtain global feature ranking. Channel selection approaches for spatial localization of the most promising brain region for drowsiness detection were incorporated to reduce intrusiveness in driving tasks. The extracted drowsiness attributes were classified using decision trees, discriminant analysis, logistic regression, naïve Bayes, support vector machines, k-nearest neighbors, and ensemble classifiers. In the comparison of all the classifiers, the ensemble model achieved the best results with 85.6% accuracy and precision, 89.7% recall, 87.6% F1-score, 70.3% Matthews correlation coefficient, 70.2% Cohen’s kappa score, and 91% area under the receiver operating characteristics curve with 76 ms execution time. These promising results are achieved at the F8 EEG channel which is selected as the channel of interest for a single channel-based drowsiness detection scheme. The results are statistically validated by a p-value of less than 0.05 with multiple significance tests. The proposed pBCI scheme demonstrates a promising technique for earlier and effective driving drowsiness detection using spectral band information of EEG biosignals.