An Interpretable Evolutionary-Fuzzy Framework for EEG Feature Extraction: Application to Chemosensory Task...

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• High-dimensional electroencephalography (EEG) data presents challenges for accurate classification.
• Existing methods often lack interpretability, hindering understanding of decision-making processes.

Purpose of the Study:

• To introduce an interpretable evolutionary-fuzzy feature extraction framework for high-dimensional EEG data.
• To automatically discover compact, nonlinear feature representations from raw EEG signals.
• To enhance classification accuracy and provide transparent decision logic.

Main Methods:

• Combined an evolution strategy (ES) optimizer with fuzzy membership encoding.
• Applied the framework to classify nasal breathing conditions during taste perception using EEG signals.
• Reduced dimensionality from 612 to 25 features.

Main Results:

• Achieved 89.50% cross-validated accuracy in distinguishing nasal breathing conditions.
• Outperformed 25-feature baselines with a 95.9% dimensionality reduction.
• Generated fully interpretable fuzzy rules for decision logic inspection.

Conclusions:

• The evolutionary-fuzzy framework offers an effective and interpretable approach for EEG classification.
• The method demonstrates robustness to noise and potential for generalization.
• Provides neuroscientists with transparent insights into EEG-based classification models.