EEG-based emotion recognition using multi-scale dynamic CNN and gated transformer.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Signal Processing

Background:

• Emotions significantly influence human cognition and decision-making.
• Electroencephalography (EEG) is a valuable tool for emotion recognition due to its temporal resolution, portability, and cost-effectiveness.

Purpose of the Study:

• To propose a novel end-to-end method, MSDCGTNet, for accurate and efficient emotion recognition from EEG signals.
• To leverage Multi-Scale Dynamic Convolutional Neural Networks (CNN) and Gated Transformer for enhanced feature extraction and dependency modeling.

Main Methods:

• Utilized Multi-Scale Dynamic CNN for extracting spatial and spectral features from raw EEG signals, minimizing information loss and computational cost.
• Employed a Gated Transformer Encoder with multi-head self-attention to capture global EEG signal dependencies efficiently.
• Applied Temporal Convolutional Network for extracting temporal features, followed by a classification module for emotion recognition.

Main Results:

• The MSDCGTNet method demonstrated high accuracy and efficiency in emotion recognition tasks.
• Evaluated on DEAP, SEED, and SEED_IV datasets, confirming the method's effectiveness.
• The approach proved robust and suitable for practical applications.

Conclusions:

• MSDCGTNet offers a valuable and effective solution for emotion recognition using EEG signals.
• The method addresses limitations of existing techniques, advancing the field of Brain-Computer Interface (BCI).
• The proposed approach enhances real-time emotion recognition capabilities.