streaming state-space rPPG with a transferable heart state
Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.
Follow-up to me-rppg by the same group (Kegang Wang, Yuntao Wang, Yuanchun Shi at Tsinghua; Daniel McDuff co-author). Same temporal-spatial state space duality idea, now with a "transferable heart state" so the network can train on long sequences and still run one frame at a time with constant memory.
temporal-spatial state space duality (TSD) SSM; transferable heart state for single-frame streaming; ~3.6 MB memory, 9.46 ms/frame (per abstract)
MIT + 'Privacy Protection Addendum' (on-device inference only; no cloud offload of video; derived metrics leave the device only with explicit consent)
Vital sign measurement using cameras presents opportunities for comfortable, ubiquitous health monitoring. Remote photoplethysmography (rPPG), a foundational technology, enables cardiac measurement through minute changes in light reflected from the skin. However, practical deployment is limited by the computational constraints of performing analysis on front-end devices and the accuracy degradation of transmitting data through compressive channels that reduce signal quality. We propose a memory efficient rPPG algorithm - FacePhys - built on temporal-spatial state space duality, which resolves the trilemma of model scalability, cross-dataset generalization, and real-time operation. Leveraging a transferable heart state, FacePhys captures subtle periodic variations across video frames while maintaining a minimal computational overhead, enabling training on extended video sequences and supporting low-latency inference. FacePhys establishes a new state-of-the-art, with a substantial 49% reduction in error. Our solution enables real-time inference with a memory footprint of 3.6 MB and per-frame latency of 9.46 ms – surpassing existing methods by 83% to 99%. These results translate into reliable real-time performance in practical deployments, and a live demo is available at https://www.facephys.com/.
Verbatim fromarXiv· arXiv metadata (abstracts are released under CC0)
Datasets linked in the hub; pretraining vs evaluation roles follow the paper.
FacePhys. (2025). arXiv preprint. https://arxiv.org/abs/2512.06275
@misc{facephys2025facephys,
title = {{FacePhys}},
year = {2025},
howpublished = {arXiv preprint},
eprint = {2512.06275},
archivePrefix = {arXiv}
}Record updated
