DiverseCER-Net: A compound emotion recognition framework based on residual Anti-Aliasing Attention network and diverse data augmentation

Key Points

• Enhanced compound emotion recognition is achieved through a novel residual network approach with anti-aliasing capabilities.
• The method shows a significant performance increase over baseline models, reaching accuracy metrics of up to 92%.
• This approach utilizes advanced data augmentation techniques that consider various emotional contexts to train the model effectively.
• Findings emphasize the need for innovative methodologies in emotion recognition, potentially impacting AI applications in mental health.