rPPG-HiBa:Hierarchical Balanced Framework for Remote Physiological Measurement

Yin Wang, Hao LU, Ying-Cong Chen, Li Kuang, Mengchu Zhou, Shuiguang Deng

Abstract:

Remote photoplethysmography (rPPG) is a promising technique for non-contact physiological signal measurement, which has great potential in health monitoring and emotion analysis. However, existing methods for the rPPG task ignore the long-tail phenomenon of physiological signal data, especially on multiple domains joint training. In addition, we find that the long-tail problem of the physiological label (phys-label) exists in different datasets, and the long-tail problem of domain exists under the same phys-label. To tackle these problems, in this paper, we propose a Hierarchical Balanced framework (rPPG-HiBa), which mitigates the bias caused by domain and phys-label imbalance. Specifically, we propose anti-spurious domain center learning tailored to learning domain-balanced embeddings space. Then, we adopt compact-aware continuity regularization to estimate phys-label-wise imbalances and construct continuity between embeddings. Extensive experiments demonstrate that our method outperforms the state-of-the-art in cross-dataset and intra-dataset settings.

Primary Subject Area: [Experience] Multimedia Applications

Secondary Subject Area: [Engagement] Emotional and Social Signals

Relevance To Conference: Our work can be a powerful tool for emotion analysis, health monitoring, future remote medical care, and intelligent driving. By extracting signals like heart rate and heart rate variability from non-intrusive facial videos, we eliminate the need for users to undergo contact-based measurements, thereby enhancing user experience and enabling a more comprehensive range of future applications. This innovative approach introduces new multimedia and multimodal processing possibilities, contributing to greater convenience and innovation in social and healthcare domains.

Submission Number: 2082