CubicPat: Investigations on the Mental Performance and Stress Detection Using EEG Signals

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Understanding brain function and its relation to mental states like stress and performance is a significant research challenge.
• Electroencephalography (EEG) signals offer a non-invasive window into brain activity.
• Existing methods often lack interpretability, hindering clinical application.

Purpose of the Study:

• To introduce a novel explainable feature engineering (XFE) architecture for analyzing EEG signals.
• To achieve accurate classification of stress and mental performance states.
• To generate interpretable results for better understanding of brain activity.

Main Methods:

• Development of a new XFE model incorporating the Cubic Pattern (CubicPat) feature extraction function.
• Utilizing cumulative weighted iterative neighborhood component analysis (CWINCA) for feature selection.
• Employing the t-algorithm-based k-nearest neighbors (tkNN) classifier for classification.
• Generating explainable results using CWINCA and Directed Lobish (DLob).

Main Results:

• The CubicPat-based XFE model achieved high classification accuracies, exceeding 95% with 10-fold cross-validation (CV) and 75% with leave-one-subject-out (LOSO) CV.
• The model demonstrated strong performance on two distinct EEG datasets for stress and mental performance detection.
• Interpretable results were successfully generated, correlating with classification outcomes.

Conclusions:

• The proposed XFE architecture effectively combines classification accuracy with interpretability for EEG-based mental state analysis.
• The CubicPat feature extraction and DLob for explainability offer a promising approach for brain signal research.
• This work contributes to solving the complexities of brain function by providing explainable insights into stress and mental performance.