EEG-Based Driving Fatigue Detection Using a Two-Level Learning Hierarchy Radial Basis Function.

Area of Science:

• Neuroscience
• Machine Learning
• Transportation Safety

Background:

• Electroencephalography (EEG) is a non-invasive, low-cost method for detecting driving fatigue.
• Extracting informative features from noisy EEG signals for fatigue detection remains a challenge.
• Radial Basis Function (RBF) neural networks offer strong non-linear approximation and generalization capabilities but face optimization challenges.

Purpose of the Study:

• To develop an enhanced RBF network, termed RBF-TLLH, for improved accuracy and efficiency in EEG-based driving fatigue detection.
• To enable global optimization of key RBF network parameters, overcoming limitations of traditional methods.
• To establish a reliable framework for detecting driver fatigue using EEG signals.

Main Methods:

• Collected experimental EEG data from participants in simulated driving environments under fatigue and alert states.
• Utilized Principal Component Analysis (PCA) for feature extraction from EEG signals.
• Employed the proposed two-level learning hierarchy RBF network (RBF-TLLH) for classification of driving status.

Main Results:

• The RBF-TLLH approach achieved a mean accuracy of 92.71% and an Area Under the ROC Curve (AUC) of 0.9199.
• Demonstrated superior classification performance compared to other widely used artificial neural networks.
• Required determination of only three core parameters, enhancing reliability and applicability.

Conclusions:

• The proposed RBF-TLLH framework offers a promising solution for reliable EEG-based driving fatigue detection.
• The method enhances classification accuracy and efficiency in identifying fatigue states.
• The simplified parameter optimization increases the practical utility of the RBF-TLLH model.