EEG-Based Driver Fatigue Detection Using FAWT and Multiboosting Approaches

Subasi, Abdulhamit; Saikia, Aditya; Bagedo, Kholoud; Singh, Amarprit; Hazarika, Anil · 2022 · Crossref

DOI: 10.1109/tii.2022.3167470

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 fatigue, which contributes to 14–20% of global road accidents. The authors propose a robust, single-modality detection system using electroencephalogram (EEG) signals to classify drivers into FATIGUE and REST states. The motivation stems from the limitations of existing methods, such as the subjectivity of self-reports, privacy concerns with video-based monitoring, and the computational complexity or "black box" nature of deep learning models. The paper aims to develop a transparent, efficient, and accurate machine learning framework suitable for real-time implementation in IoT and automotive systems. The methodology utilizes an open-source EEG dataset collected from sixteen participants using a driving simulator. EEG signals were preprocessed to remove noise and artifacts, then decomposed using Flexible Analytic Wavelet Transform (FAWT). FAWT was selected for its ability to handle non-stationary signals and provide flexible time-frequency coverage. From the FAWT sub-bands, ninety-five low-order statistical features (including mean, standard deviation, skewness, and kurtosis) were extracted. These features underwent Linear Discriminant Analysis (LDA) for dimensionality reduction and statistical significance testing (p < 0.05) to retain only the most discriminative measures. The final classification was performed using a Multiboosting ensemble strategy, which combines weak learners (such as SVM, ANN, and Rotation Forest) to reduce variance and overfitting. The model was validated using 10-fold cross-validation. The results demonstrate that the proposed FAWT-based Multiboosting approach achieves high classification performance. The optimal model, Multiboost-SVM, attained an accuracy of 97.10% for the FATIGUE state and 97.90% for the REST state. It also achieved an F-score of 97.50%, an Area Under the Curve (AUC) of 0.975, and a Kappa coefficient of 0.950. Other Multiboost variants, including Multiboost-ANN and Multiboost-Rotation Forest, also showed strong performance with accuracies above 96%. The study notes that the REST state was generally easier to classify than the FATIGUE state, likely due to higher discriminant feature separability. The computational time for feature extraction and classification was reported to be within a range suitable for real-time applications. The significance of this work lies in its demonstration that a single-modality EEG approach, combined with advanced signal processing and ensemble learning, can rival or exceed multimodal systems in accuracy while maintaining interpretability and lower computational complexity. By avoiding the pitfalls of deep learning black boxes and the constraints of video-based monitoring, this method offers a reliable solution for real-time driver fatigue detection. The high performance metrics suggest the system is viable for integration into smart vehicles and IoT platforms to enhance road safety by alerting drivers before fatigue leads to 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 qwen3.6-27b-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