Functional Connectivity Analysis in Multi-channel EEG for Emotion Detection with Phase Locking Value and 3D CNN.

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electroencephalogram (EEG) signal analysis is crucial for understanding brain activity.
• Accurate emotion detection from EEG remains a challenge, requiring advanced signal processing and machine learning techniques.

Purpose of the Study:

• To develop and evaluate a novel system for emotion detection using EEG signals.
• To leverage noise-assisted multivariate Empirical Mode Decomposition (NA-MEMD) for enhanced feature extraction.
• To utilize 3D convolutional neural networks (CNNs) for spatial-temporal feature learning and classification.

Main Methods:

• Applied NA-MEMD to multi-channel EEG signals to extract intrinsic mode functions (IMFs).
• Performed functional connectivity analysis using phase locking value (PLV) to create connectivity maps.
• Employed a 3D CNN to learn spatial-temporal features from connectivity maps for emotion classification.

Main Results:

• The system achieved high accuracy in binary emotion classification (valence: 97.37%, arousal: 96.26%).
• Multi-class emotion classification accuracy reached 94.78% on the DEAP dataset and 99.54% on the SEED dataset.
• The proposed system outperformed existing deep learning models and conventional EEG feature-based methods.

Conclusions:

• The proposed NA-MEMD and 3D CNN-based system demonstrates superior performance for EEG-based emotion detection.
• This approach effectively captures spatial-temporal brain activity patterns for robust emotion classification.
• The findings suggest a promising direction for developing advanced brain-computer interfaces for emotion recognition.