Improved drowsiness detection in drivers through optimum pairing of EEG features using an optimal EEG channel comparable to a multichannel EEG system

Minhas, Riaz; Peker, Nur Yasin; Hakkoz, Mustafa Abdullah; Arbatli, Semih; Celik, Yeliz; Erdem, Cigdem Eroglu; Peker, Yuksel; Semiz, Beren · 2025 · Crossref

DOI: 10.1007/s11517-025-03375-1

archive: archived pipeline: cataloged verified

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

Summary

This study addresses the trade-off between the high accuracy of multichannel electroencephalography (EEG) systems and the practical limitations of single-channel devices for driver drowsiness detection. While multichannel setups offer superior coverage, they are computationally demanding and uncomfortable for users. Conversely, single-channel devices are user-friendly but typically provide lower coverage. The authors hypothesized that an optimal single EEG channel, paired with specific EEG features, could achieve coverage comparable to a multichannel system, thereby reducing hardware complexity without sacrificing performance. The research utilized data from 50 professional drivers diagnosed with obstructive sleep apnea who participated in a 50-minute simulated driving session. Drowsiness ground truth was established using visual-based scoring derived from facial video analysis, specifically integrating PERCLOS (percentage of eyelid closure) and eyelid closure duration metrics. This yielded 927 EEG epochs (453 drowsy, 474 wakeful). EEG signals were recorded from six channels (F3, F4, C3, C4, O1, O2) and processed using discrete wavelet transform (DWT) to extract ten normalized features, including power spectral density (PSD) for alpha, theta, and delta bands, and their ratios. The study evaluated seven subject-specific and feature-specific thresholding techniques to classify epochs. To identify the optimal configuration, the authors assessed 45 combinations of paired EEG features across the six channels, comparing their coverage and accuracy against the visual-based scoring reference. The results demonstrated that pairing PSD alpha and PSD theta features in the Frontal4 (F4) and Occipital2 (O2) channels yielded the highest performance. The F4 channel achieved 96.1% coverage and 95.4% accuracy, while the O2 channel achieved 95% coverage and 94.7% accuracy. These single-channel results slightly surpassed the coverage achieved by a six-channel system using a single feature, with improvements of 1.47% for F4 and 0.32% for O2. The study identified F4 as the optimal channel due to its superior sensitivity in detecting drowsiness transitions. The significance of this work lies in its demonstration that reducing EEG channels from six to one is feasible without compromising detection efficacy. By identifying an optimal channel and feature pair, the approach lowers computational demands and hardware requirements, making wearable drowsiness detection systems more practical for real-world applications. The findings provide a validated methodology for developing cost-effective, single-channel EEG devices that maintain high reliability in monitoring driver alertness.

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 qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

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

• drowsiness detection algorithms
• microsleep
• drowsiness

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
• Methodological Resource: validation psychometrics