Development of Drowsy Driving Detection System Using EEG

Seo, Ssang-Hee · 2026 · Crossref

DOI: 10.31803/tg-20250326030355

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 drowsy driving, a leading cause of traffic accidents with high fatality rates. Motivated by the limitations of existing detection methods and the need for real-time, efficient monitoring, the authors propose a lightweight electroencephalogram (EEG)-based drowsiness detection system. The system integrates EEG signal acquisition, preprocessing, feature extraction, and classification using a custom Convolutional Neural Network (CNN), coupled with an Android application for immediate user feedback. The experimental design involved four healthy adult male participants (average age 24.5 years) who viewed a 20-minute highway driving video while wearing a DSI-24 wireless EEG headset. Data were collected from 21 sensors, with specific focus on the prefrontal (Fp1, Fp2) and occipital (O1, O2) regions. Preprocessing included band-pass filtering (0–50 Hz) and Independent Component Analysis to remove artifacts. Feature extraction focused on the relative power of the Alpha frequency band (8–13 Hz), identified as the most discriminative metric between wakeful and drowsy states. The processed features were transmitted via the OpenViBE platform and Lab Streaming Layer to a Flask server, where the lightweight CNN model performed real-time classification. The model architecture consisted of three convolutional layers and two fully connected layers, optimized for low computational overhead. The proposed CNN achieved a classification accuracy of 91.56%, with a precision of 0.9476 and recall of 0.9272. This performance was comparable to the more complex AlexNet model, which achieved 92.66% accuracy, but the proposed model required significantly fewer parameters, making it suitable for embedded systems. The system’s effectiveness was further validated by comparing input types; using extracted Alpha relative power features yielded 91.56% accuracy, whereas raw EEG data resulted in only 62.17% accuracy. The Android application successfully provided real-time visual and auditory warnings and recommended nearby rest areas upon detecting drowsiness. The study concludes that a lightweight CNN utilizing specific EEG features offers a practical, efficient solution for real-time drowsy driving detection. By balancing high accuracy with low computational cost, the system is viable for deployment in resource-constrained environments. However, the authors acknowledge limitations regarding the small sample size and controlled experimental conditions. Future work will focus on large-scale data collection under actual driving conditions to further validate and improve the system’s robustness and generalizability.

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
• drowsiness
• microsleep

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