ARGen: Affect-Reinforced Generative Augmentation towards Vision-based Dynamic Emotion Perception
Summary: arXiv:2604.12255v1 Announce Type: cross
Abstract
Dynamic facial expression recognition in the wild remains challenging due to data scarcity and long-tail distributions, which hinder models from effectively learning the temporal dynamics of scarce emotions. To address these limitations, we propose ARGen, an Affect-Reinforced Generative Augmentation Framework that enables data-adaptive dynamic expression generation for robust emotion perception.
Framework Overview
ARGen operates in two main stages:
- Affective Semantic Injection (ASI): This stage establishes affective knowledge alignment through facial Action Units. It employs a retrieval-augmented prompt generation strategy to synthesize consistent and fine-grained affective descriptions via large-scale visual-language models, thereby injecting interpretable emotional priors into the generation process.
- Adaptive Reinforcement Diffusion (ARD): This stage integrates text-conditioned image-to-video diffusion with reinforcement learning. It introduces inter-frame conditional guidance and a multi-objective reward function to jointly optimize expression naturalness, facial integrity, and generative efficiency.
Key Contributions
ARGen presents several key contributions to the field of dynamic emotion perception:
- Development of a novel framework that adapts to data scarcity and enhances model learning of temporal dynamics.
- Utilization of Affective Semantic Injection to improve the interpretability and accuracy of generated emotional expressions.
- Implementation of Adaptive Reinforcement Diffusion, which effectively combines textual and visual data to create high-quality video outputs that reflect authentic emotional states.
Experimental Validation
Extensive experiments conducted on both generation and recognition tasks demonstrate that ARGen substantially enhances synthesis fidelity and improves recognition performance. The results indicate that ARGen establishes an interpretable and generalizable generative augmentation paradigm for vision-based affective computing.
Conclusion
ARGen stands as a significant advancement in the realm of emotion recognition technology, addressing the critical challenges posed by data limitations and the complexities of dynamic facial expressions. Its innovative approach offers a promising path toward more robust and effective emotion perception systems, paving the way for future research in affective computing.