CLDAE: A Two Stage EEG-based Emotion Recognition Framework Combining Contrastive Learning and Dual-Attention Encoder.

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Electroencephalogram (EEG)-based emotion recognition faces challenges in cross-subject generalization and within-subject personalization.
• Existing models often require extensive personalized data or struggle to adapt to individual neural signatures.

Purpose of the Study:

• To propose a novel EEG-based emotion recognition framework, CLDAE, addressing generalization and personalization limitations.
• To integrate contrastive learning and a dual-attention feature extraction mechanism for improved performance.

Main Methods:

• CLDAE framework involves two stages: contrastive learning pre-training and emotion recognition fine-tuning.
• Utilizes data augmentation with cross-subject EEG signals for pre-training.
• Employs a dual-attention encoder combining temporal and channel attention for feature extraction.

Main Results:

• CLDAE achieved competitive performance on both within-subject and cross-subject emotion recognition tasks.
• Demonstrated high accuracy on the MAN dataset: 95.12% for within-subject and 75.29% for cross-subject recognition.
• Outperformed baseline methods in emotion recognition accuracy.

Conclusions:

• The proposed CLDAE framework effectively enhances EEG-based emotion recognition.
• The integration of contrastive learning and dual-attention mechanisms proves beneficial for both personalization and generalization.