Toward Practical Driver Fatigue Detection Based on EEG Using Forehead Low Channel and Cascaded Deep Forest

Min, Jianliang; Qiu, Bo · 2025 · Crossref

DOI: 10.21203/rs.3.rs-7075565/v1

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 challenge of developing practical, real-time driver fatigue detection systems using electroencephalogram (EEG) signals. While EEG is a reliable physiological indicator for fatigue, traditional multi-channel headsets are impractical for everyday driving due to their complexity and discomfort. The authors propose a solution utilizing low-channel forehead EEG, which is compatible with portable wearable devices, combined with a Cascaded Deep Forest (DF) classification framework. The primary objective is to create an efficient, robust, and accurate detection method that overcomes the limitations of image-based techniques and complex multi-channel setups. The experimental design involved 26 healthy subjects participating in a simulated driving task lasting nearly two hours. EEG data were recorded using a Neuroscan system, with preprocessing focused on the frontal channels Fp1 and Fp2. To ensure high-confidence labeling of fatigue states, the researchers employed a comprehensive evaluation metric combining three scores: facial expression analysis (including PERCLOS and yawning), driving performance errors, and subjective questionnaire responses. Feature extraction utilized two specific methods: Wavelet Log-Energy Entropy (WLE) to capture energy information across different frequency bands, and Component Statistical Features (CSF) derived from reconstructed phase space analysis to characterize nonlinear dynamic states. These features were fed into a cascaded deep forest model, which integrates random forests and extra trees to enhance robustness and generalization. The results demonstrated that the proposed DF model achieved an average accuracy of 95.1%, with sensitivity of 94.9%, specificity of 95.3%, and a Matthews correlation coefficient of 90.2%. Statistical tests confirmed that this performance was significantly superior to classical classifiers, including Logistic Regression, Support Vector Machines, Random Forests, and LightGBM. The study also highlighted the computational efficiency of the feature extraction process, with calculations completed in under 0.005 seconds per epoch, supporting real-time application. Furthermore, analysis revealed that combining both Fp1 and Fp2 channels significantly improved detection accuracy compared to using either single channel alone, which yielded accuracies below 92%. The significance of this work lies in its contribution to the development of wearable, low-channel EEG systems for traffic safety. By demonstrating that forehead EEG combined with efficient feature extraction and deep forest classification can achieve high accuracy, the study provides a viable pathway for integrating fatigue detection into consumer-grade wearable devices. The findings suggest that this approach offers a robust alternative to multi-channel systems, facilitating the practical deployment of real-time fatigue monitoring in driving environments.

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

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