Driving drowsiness detection using spectral signatures of EEG-based neurophysiology

Abstract

Introduction: Drowsy driving is a significant factor causing dire road crashes and casualties around the world. Detecting it earlier and more effectively can significantly reduce the lethal aftereffects and increase road safety. As physiological conditions originate from the human brain, so neurophysiological signatures in drowsy and alert states may be investigated for this purpose. In this preface, A passive brain-computer interface (pBCI) scheme using multichannel electroencephalography (EEG) brain signals is developed for spatially localized and accurate detection of human drowsiness during driving tasks. Methods: This pBCI modality acquired electrophysiological patterns of 12 healthy subjects from the prefrontal (PFC), frontal (FC), and occipital cortices (OC) of the brain. Neurological states are recorded using six EEG channels spread over the right and left hemispheres in the PFC, FC, and OC of the sleep-deprived subjects during simulated driving tasks. In post-hoc analysis, spectral signatures of the δ, θ, α, and β rhythms are extracted in terms of spectral band powers and their ratios with a temporal correlation over the complete span of the experiment. Minimum redundancy maximum relevance, Chi-square, and ReliefF feature selection methods are used and aggregated with a Z-score based approach for global feature ranking. The extracted drowsiness attributes are classified using decision trees, discriminant analysis, logistic regression, naïve Bayes, support vector machines, k-nearest neighbors, and ensemble classifiers. The binary classification results are reported with confusion matrix-based performance assessment metrics. Results: In inter-classifier comparison, the optimized ensemble model achieved the best results of drowsiness classification with 85.6% accuracy and precision, 89.7% recall, 87.6% F₁-score, 80% specificity, 70.3% Matthews correlation coefficient, 70.2% Cohen's kappa score, and 91% area under the receiver operating characteristic curve with 76-ms execution time. In inter-channel comparison, the best results were obtained at the F8 electrode position in the right FC of the brain. The significance of all the results was validated with a p-value of less than 0.05 using statistical hypothesis testing methods. Conclusions: The proposed scheme has achieved better results for driving drowsiness detection with the accomplishment of multiple objectives. The predictor importance approach has reduced the feature extraction cost and computational complexity is minimized with the use of conventional machine learning classifiers resulting in low-cost hardware and software requirements. The channel selection approach has spatially localized the most promising brain region for drowsiness detection with only a single EEG channel (F8) which reduces the physical intrusiveness in normal driving operation. This pBCI scheme has a good potential for practical applications requiring earlier, more accurate, and less disruptive drowsiness detection using the spectral information of EEG biosignals.

Keywords: brain–computer interface; channel selection; drowsiness detection; electroencephalography; feature selection; neurophysiology; spectral features; supervised learning.

FIGURE 1

(A) Process flow diagram for BCI module. (B) Experiment details.

FIGURE 2

(A) OpenBCI Ultracortex Mark-IV EEG headset in a 16-channel configuration. (B) EEG referential montage according to the 10–20 system of electrode placement with selected channels highlighted in red.

FIGURE 3

2D scatter plots on a log scale for feature pairs between band power ratios (top three rows), and spectral band powers (bottom three rows). Fp1, Fp2, O1, O2, F7, and F8 channels in each row from top to bottom, respectively. Red and blue data points represent drowsy and alert states, respectively.

FIGURE 4

3D scatter plots on a log scale for all the ternary combinations of five top-ranked features. Red and blue data points represent drowsy and alert states, respectively (channel F8, all subjects).

FIGURE 5

Welch’s power spectral density estimates for selected PFC, FC, and OC channels over the spectral range of four EEG bands (subjects 1–5).

FIGURE 6

Spectrograms for the complete experiment over the frequency range of all EEG bands for the selected channels (subject 2).

FIGURE 7

Spectrograms showing spectral band powers in individual EEG frequency bands in the F8 channel (subject 4).

FIGURE 8

Channel-wise feature ranking with MRMR, chi-square, and ReliefF methods with predictor importance scores over all subjects’ data. The cumulative effect of all feature selection methods is shown with a global feature ranking scheme with predictor importance in terms of Z-score.

FIGURE 9

(A) Classification performance comparison among feature combinations of spectral feature space for the F8 channel with ensemble classifier over all subject’s data. (B) Classification performance comparison among various classifiers for the F8 channel over all subject’s data with variance bounds obtained with varying feature combinations.

FIGURE 10

(A) Heat map chart for performance comparison with percentage accuracy of all the classification methods using $\delta ,\theta ,\alpha ,\beta$ band powers only and channel ranking for the best classifier. (B) Classification computation time (ms) for all the classifiers mentioning their computational complexity for driving drowsiness assessment.

FIGURE 11

(A) Classification results in terms of confusion matrices. (B) Receiver operating characteristics curves over all subjects’ data in each selected channel with Bayesian optimization-based ensemble classifier. Class labels: 1 = drowsy, 0 = alert.