BTS-rPPG: Orthogonal Butterfly Temporal Shifting for Remote Photoplethysmography

TL;DR

BTS-rPPG introduces an innovative butterfly-inspired temporal shifting framework with orthogonal feature transfer to enhance long-range temporal modeling in remote photoplethysmography from facial videos.

Contribution

The paper proposes BTS-rPPG, a novel structured temporal modeling method based on FFT-inspired butterfly communication and orthogonal feature transfer for improved rPPG signals.

Findings

Outperforms existing methods on multiple benchmark datasets.

Enhances long-range temporal modeling of physiological signals.

Efficiently propagates information across distant frames.

Abstract

Remote photoplethysmography (rPPG) enables contactless physiological sensing from facial videos by analyzing subtle appearance variations induced by blood circulation. However, modeling the temporal dynamics of these signals remains challenging, as many deep learning methods rely on temporal shifting or convolutional operators that aggregate information primarily from neighboring frames, resulting in predominantly local temporal modeling and limited temporal receptive fields. To address this limitation, we propose BTS-rPPG, a temporal modeling framework based on Orthogonal Butterfly Temporal Shifting (BTS). Inspired by the butterfly communication pattern in the Fast Fourier Transform (FFT), BTS establishes structured frame interactions via an XOR-based butterfly pairing schedule, progressively expanding the temporal receptive field and enabling efficient propagation of information across distant frames. Furthermore, we introduce an orthogonal feature transfer mechanism (OFT) that filters the source feature with respect to the target context before temporal shifting, retaining only the orthogonal component for cross-frame transmission. This reduces redundant feature propagation and encourages complementary temporal interaction. Extensive experiments on multiple benchmark datasets demonstrate that BTS-rPPG improves long-range temporal modeling of physiological dynamics and consistently outperforms existing temporal modeling strategies for rPPG estimation.