FreqPhys: Repurposing Implicit Physiological Frequency Prior for Robust Remote Photoplethysmography

2026-04-01

AI Summary

This work addresses the limited robustness of existing remote photoplethysmography (rPPG) methods under motion artifacts and illumination variations, which predominantly rely on time-domain modeling. To overcome this, we propose FreqPhys, a novel framework that explicitly incorporates and reuses physiological frequency priors. FreqPhys integrates physiological bandpass filtering, spectral modulation with adaptive selection, cross-domain representation learning, and a time-frequency fused conditional diffusion mechanism to effectively synergize frequency-domain guidance with deep time-domain features. Extensive experiments on six benchmark datasets demonstrate that our method significantly outperforms state-of-the-art approaches, particularly excelling in scenarios with severe motion interference. These results underscore the critical role of explicit frequency-domain prior modeling in enhancing the robustness of rPPG signal recovery.

Abstract

Remote photoplethysmography (rPPG) enables contactless physiological monitoring by capturing subtle skin-color variations from facial videos. However, most existing methods predominantly rely on time-domain modeling, making them vulnerable to motion artifacts and illumination fluctuations, where weak physiological clues are easily overwhelmed by noise. To address these challenges, we propose FreqPhys, a frequency-guided rPPG framework that explicitly leverages physiological frequency priors for robust signal recovery. Specifically, FreqPhys first applies a Physiological Bandpass Filtering module to suppress out-of-band interference, and then performs Physiological Spectrum Modulation together with adaptive spectral selection to emphasize pulse-related frequency components while suppress residual in-band noise. A Cross-domain Representation Learning module further fuses these spectral priors with deep time-domain features to capture informative spatial--temporal dependencies. Finally, a frequency-aware conditional diffusion process progressively reconstructs high-fidelity rPPG signals. Extensive experiments on six benchmarks demonstrate that FreqPhys yields significant improvements over state-of-the-art approaches, particularly under challenging motion conditions. It highlights the importance of explicitly modeling physiological frequency priors. The source code will be released.

Problem

• remote photoplethysmography
• motion artifacts
• illumination fluctuations
• physiological signal recovery
• noise robustness

Innovation

• frequency prior
• remote photoplethysmography
• spectral modulation
• conditional diffusion
• cross-domain representation