Deep Hybrid CNN-BiLSTM-Attention Model for EEG Classification Using Wavelet Features

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Electroencephalography (EEG) offers excellent temporal resolution for studying brain dynamics.
• Class imbalance and variability in EEG data pose challenges for automatic brain state discrimination.
• Distinguishing cognitive states from EEG requires advanced analytical methods.

Purpose of the Study:

• To categorize EEG recordings into distinct brain states (rest vs. auditory mantra stimulation).
• To improve the discriminative learning from wavelet-based time-frequency features using a deep hybrid neural network.
• To enhance the accuracy of automatic brain state classification.

Main Methods:

• EEG data recorded from experienced practitioners in resting and mantra-listening states.
• Wavelet transforms used for time-frequency representation of EEG segments.
• A hybrid deep learning model combining Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and attention mechanisms was employed.

Main Results:

• The proposed CNN-BiLSTM-Attention model achieved a high accuracy of 99.46% on an independent test set.
• This significantly outperformed baseline models like CNN, LSTM, and CNN+LSTM.
• Receiver Operating Characteristic (ROC) analysis showed an Area Under the Curve (AUC) near 1.0, confirming strong discriminative capability.

Conclusions:

• The hybrid deep learning framework effectively enhances spatial-temporal feature learning for EEG analysis.
• The model demonstrates robust performance in distinguishing between resting and during-mantra auditory stimulation brain states.
• This approach shows potential for neurophysiological monitoring and real-time cognitive state detection.