Classification of driver fatigue in an electroencephalography-based countermeasure system with source separation...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Driver fatigue is a major cause of road accidents.
• Electroencephalography (EEG) offers a potential method for objective fatigue detection.
• Accurate classification of fatigue states is crucial for developing effective countermeasures.

Purpose of the Study:

• To classify driver fatigue and alert states using EEG signals.
• To evaluate the effectiveness of a novel source separation technique (ICA-ERBM) in improving fatigue classification accuracy.
• To compare classification performance with and without the source separation module.

Main Methods:

• Utilized electroencephalography (EEG) data from 43 participants.
• Employed power spectral density (PSD) for feature extraction.
• Implemented a fuzzy swarm-based artificial neural network (ANN) for classification.
• Investigated Independent Component Analysis of Entropy Rate Bound Minimization (ICA-ERBM) as a source separation technique.

Main Results:

• Without source separation, classification accuracy was 73.65% (sensitivity 71.67%, specificity 75.63%).
• With the inclusion of the ICA-ERBM source separator, classification accuracy significantly improved to 78.88% (sensitivity 78.16%, specificity 79.60%; p < 0.05).
• The source separation module demonstrated a clear benefit in distinguishing between fatigue and alert states.

Conclusions:

• The proposed EEG-based system with ICA-ERBM source separation is effective for driver fatigue detection.
• Source separation techniques can enhance the performance of EEG-based fatigue classification systems.
• This approach holds promise for developing advanced driver assistance systems to improve road safety.