Driver Drowsiness Detection with Commercial EEG Headsets

Delrobaei, Mehdi; Rezaee, Qazal · 2023 · Crossref

DOI: 10.31224/2933

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 critical safety issue of driver drowsiness, a leading cause of road accidents, by evaluating the efficacy of commercial electroencephalography (EEG) headsets compared to traditional vehicle-based detection methods. While vehicle-based metrics (e.g., steering angle, lane deviation) are non-intrusive and easy to capture, they often suffer from low accuracy due to external disturbances and delayed physiological responses. Conversely, clinical EEG methods offer high accuracy but are impractical for real-world driving due to their intrusive nature. The authors investigate whether consumer-grade dry-electrode EEG headsets can provide a feasible, high-performance alternative for detecting drowsiness in real-time driving scenarios. The experimental design involved 50 volunteers, with data from 48 subjects analyzed, who drove a fixed-base simulator under sleep-deprived conditions (averaging 4.5 hours of sleep the previous night). Participants wore commercial Muse 2 or Muse S headsets, which recorded EEG signals from four channels (AF7, AF8, TP9, TP10) at 256 Hz. Drowsiness levels were labeled using the Observer Rating of Drowsiness (ORD) method, where three observers rated video footage of the drivers every 30 seconds on a scale of 1 to 5. Ratings of 1–2 were classified as alert, and 3–5 as drowsy. The study compared the statistical separation of alert and drowsy states using EEG features (absolute and relative power spectral density across delta, theta, alpha, beta, and gamma bands) against vehicle-based features (steer angle, steer speed, lane deviation, and torque sensor). The results demonstrated that EEG-based features provided significantly better separation between alert and drowsy states than vehicle-based features. Statistical analysis using Wilcoxon rank-sum tests revealed that EEG features, particularly absolute power spectral density, yielded much lower p-values, indicating stronger statistical significance. Specifically, beta, theta, and gamma frequency bands were most effective in distinguishing drowsiness. In contrast, vehicle-based metrics showed weaker performance; while steer angle, lane deviation, and torque sensor values were statistically significant, steer speed was not (p > 0.05). Furthermore, 40% of relative PSD EEG features failed to show significant separation, whereas only 5% of absolute PSD features did, suggesting absolute PSD is the more robust metric for this application. The study concludes that commercial EEG headsets are viable and superior alternatives to vehicle-based methods for driver drowsiness detection. By leveraging consumer-grade devices, the approach mitigates the complexity and intrusiveness of clinical EEG systems while maintaining high detection accuracy. This finding supports the integration of affordable, wearable EEG technology into driver monitoring systems to enhance road safety by providing earlier and more reliable drowsiness alerts than current vehicle-based sensors.

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_with_issues.

Topics

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

• drowsiness detection algorithms
• drowsiness
• microsleep
• dms validation

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: tool software, validation psychometrics