Driving drowsiness detection using spectral signatures of EEG-based neurophysiology

Arif, Saad; Munawar, Saba; Ali, Hashim · 2023 · Crossref

DOI: 10.3389/fphys.2023.1153268

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical safety issue of drowsy driving by developing a passive brain-computer interface (pBCI) scheme for the early and accurate detection of driver drowsiness using electroencephalography (EEG). The research is motivated by the high incidence of fatal accidents caused by fatigue and the need for detection methods that are less intrusive than behavioral or vehicular monitoring systems. The authors hypothesize that neurophysiological signatures in specific brain regions can reliably distinguish between alert and drowsy states, allowing for a system that minimizes hardware complexity and driver discomfort. The experimental design involved 12 healthy, right-handed male subjects (average age 30 ± 2 years) who were sleep-deprived and performed simulated driving tasks. EEG data were acquired using a 16-channel OpenBCI headset, with six specific channels selected for analysis: two each from the prefrontal cortex (Fp1, Fp2), frontal cortex (F7, F8), and occipital cortex (O1, O2). Signal processing included Gaussian filtering, notch rejection for 50/60 Hz interference, and a low-pass Butterworth filter to retain the 0.5–40 Hz range. Spectral features were extracted using Welch’s Power Spectral Density (PSD) and spectrograms, focusing on delta, theta, alpha, and beta band powers and their ratios. Feature selection was performed using Minimum Redundancy Maximum Relevance, Chi-square, and ReliefF methods, aggregated via Z-score ranking. Classification was conducted using decision trees, discriminant analysis, logistic regression, naïve Bayes, support vector machines, k-nearest neighbors, and ensemble classifiers. The results demonstrated that an optimized ensemble model achieved the highest performance, with 85.6% accuracy and precision, 89.7% recall, and an area under the receiver operating characteristic curve of 91%. The model operated with a low execution time of 76 ms. Statistical hypothesis testing confirmed the significance of these results (p < 0.05). A key finding from the inter-channel comparison was that the F8 electrode position in the right frontal cortex yielded the best detection results. This spatial localization allowed the system to function effectively using a single EEG channel, significantly reducing the physical intrusiveness of the monitoring setup. The significance of this work lies in its practical applicability for road safety. By identifying the F8 channel as the most promising region for drowsiness detection, the study enables the development of low-cost, non-intrusive hardware solutions. The use of conventional machine learning classifiers rather than deep learning reduces computational complexity and data requirements, making the system more feasible for real-world vehicular integration. The approach successfully balances detection accuracy with minimal driver disruption, offering a viable pathway for earlier intervention in fatigue-related accidents.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• drowsiness
• microsleep
• neuro workload indices

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data