DC-ASTGCN: EEG Emotion Recognition Based on Fusion Deep Convolutional and Adaptive Spatio-Temporal Graph...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Growing interest in emotion recognition using electroencephalogram (EEG) signals, driven by advancements in artificial intelligence (AI) and brain-computer interfaces (BCI).
• Challenges in accurately classifying emotions from complex EEG data due to the need to integrate time, frequency, and spatial domain features.

Purpose of the Study:

• To propose a novel fusion model, DC-ASTGCN, for comprehensive analysis and understanding of EEG signals for emotion recognition.
• To address the limitations of existing methods in capturing the intricate features of EEG data for emotion classification.

Main Methods:

• Developed DC-ASTGCN, a hybrid model combining a deep convolutional neural network (DCNN) for frequency and local spatial feature extraction.
• Integrated an adaptive spatio-temporal graphic convolutional network (ASTGCN) with attention mechanisms to analyze functional connectivity between brain regions.
• Utilized DCNN to identify brain region activity patterns and ASTGCN to reveal inter-regional functional connectivity in various emotional states.

Main Results:

• The DC-ASTGCN model demonstrated superior performance in emotion recognition accuracy compared to existing state-of-the-art methods.
• Experiments on the DEAP and SEED datasets validated the effectiveness of the proposed fusion model.
• The model successfully integrated diverse EEG features for enhanced emotional state understanding.

Conclusions:

• The DC-ASTGCN model offers a significant advancement in EEG-based emotion recognition by effectively integrating DCNN and ASTGCN.
• The proposed approach enhances the ability to understand and recognize emotional states by comprehensively analyzing complex EEG signals.
• This research contributes to the field of affective computing and BCI applications.