Optimizing deep CNN architecture via hybrid Harris Hawks arithmetic algorithm for EEG meditation classification.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Signal Processing

Background:

• Meditation enhances cognitive function, but analyzing EEG signals for classification remains challenging.
• Existing methods use limited features and traditional machine learning, lacking advanced techniques.
• There's a need for integrated approaches combining time-frequency analysis, deep learning, and optimization for EEG meditation classification.

Purpose of the Study:

• To develop a hybrid EEG-based framework for classifying meditation states.
• To enhance meditation classification accuracy by integrating advanced optimization and deep learning techniques.
• To address limitations in current EEG signal analysis for meditation research.

Main Methods:

• A hybrid framework combining Harris Hawks Optimization (HHO) and Arithmetic Optimization Algorithm (AOA) was developed to tune Convolutional Neural Network (CNN) parameters.
• EEG signals were pre-processed and transformed into time-frequency images using the Stockwell Transform (S-transform).
• The HHO-AOA-CNN model processed these images for hyper-parameter optimization and classification of Vipassana (VIP), Isha Shoonya (IS), and Control (CTR) states.

Main Results:

• The proposed HHO-AOA-CNN framework achieved a classification accuracy of 94.20%.
• The hybrid model demonstrated superior performance compared to standalone HHO-CNN, AOA-CNN, and baseline CNN models.
• Statistical analysis confirmed the stability and robustness of the hybrid optimization approach.

Conclusions:

• The developed HHO-AOA-CNN framework offers a robust and accurate method for EEG-based meditation classification.
• This approach effectively integrates advanced signal processing, deep learning, and optimization techniques.
• The findings contribute to improved understanding and objective measurement of meditation states through EEG analysis.