Explainable spatial-temporal-spectral meta-learning for subject-independent EEG-based emotion recognition

Key Points

• The emotion recognition accuracy increased, achieving a performance improvement of 15% over previous models.
• Key findings include enhanced interpretability and robustness using meta-learning techniques applied to EEG signals.
• Observational analysis focused on spatial, temporal, and spectral features gathered from EEG recordings across various conditions.
• This approach may enable better emotion recognition systems, but further validation in real-world scenarios is needed.