pith. sign in

arxiv: 2605.17859 · v2 · pith:PQW3AH6Xnew · submitted 2026-05-18 · 💻 cs.HC · cs.LG

Multi-site PPG: An In-the-Wild Physiological Dataset from Emerging Multi-site Wearables

Jiayi Shao , Jiaying Ye , Shengyao Liu , Zachary Englhardt , Girish Narayanswamy , Vikram Iyer , Qiuyue Shirley Xue This is my paper

Pith reviewed 2026-05-20 09:29 UTC · model grok-4.3

classification 💻 cs.HC cs.LG
keywords PPGphotoplethysmographywearableheart ratedatasetmulti-sitein-the-wildphysiological sensing
0
0 comments X

The pith

Heart rate estimation from PPG works best on earrings with 2.30 bpm error and worst on necklaces with 8.68 bpm error in everyday settings.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper releases the Multi-site PPG dataset gathered from four custom wearables—an earring, ring, watch, and necklace—worn during normal daily activities. The data includes green and infrared PPG, acceleration, temperature, and reference ECG, with over 350 hours of recordings and hundreds of hours of ready-to-use windows. Benchmarking heart rate methods on the dataset uncovers large performance gaps tied to body site. The analysis also covers motion impacts and benefits from multi-site or PPG-acceleration fusion approaches.

Core claim

The authors created a dataset of simultaneous multi-site PPG recordings from emerging wearable form factors including a smart earring, ring, watch, and necklace. Participants wore these devices along with a chest ECG reference while going about their normal routines for multiple days. The resulting data allows direct comparison of heart rate estimation algorithms, which achieve the lowest mean absolute error of 2.30 bpm at the earring site and the highest of 8.68 bpm at the necklace site.

What carries the argument

The key mechanism is the synchronized multi-site PPG data collection from different body locations with a shared ECG ground truth to quantify site-specific differences in physiological signal quality.

If this is right

  • Accuracy of PPG-based heart rate tracking depends heavily on which body site the wearable is placed at.
  • Earrings provide the most accurate readings among the four tested form factors for in-the-wild use.
  • Motion artifacts affect different sites to varying degrees, suggesting tailored processing per device.
  • Fusing signals from multiple sites or with acceleration data offers a path to more reliable estimates.
  • The dataset can guide the design of future wearables optimized for specific placements.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Device makers may focus development on earring-style wearables for superior health monitoring performance.
  • The findings could extend to other metrics like respiration rate or blood pressure estimation from PPG.
  • Researchers can use this data to develop adaptive algorithms that switch between sites based on signal quality.
  • Similar site variations might appear in other sensing modalities such as temperature or bioimpedance.

Load-bearing premise

The custom devices maintain consistent skin contact and optical alignment during unconstrained daily activities without systematic placement or motion artifacts that would invalidate cross-site comparisons.

What would settle it

If a new collection of data from the same sites but with enforced uniform motion and contact conditions shows comparable mean absolute errors across all four wearables, the site-specific performance differences would be called into question.

Figures

Figures reproduced from arXiv: 2605.17859 by Girish Narayanswamy, Jiaying Ye, Jiayi Shao, Qiuyue Shirley Xue, Shengyao Liu, Vikram Iyer, Zachary Englhardt.

Figure 1
Figure 1. Figure 1: Multi-site data collection setup. Our dataset collects signals from four wearable devices: earring, ring, watch, necklace. Each device shares the same sensing hardware and streams two-channel reflective PPG (green and IR), 3-axis accelerometer, and temperature data to a smartphone app. A Polar H10 chest strap provides reference ECG for ground truth. to quantify the benefits of non-wrist sensing, characteri… view at source ↗
Figure 2
Figure 2. Figure 2: (a) Hardware design of the wearable sensing platform. Each device is built on the same PCB integrating the microcontroller and PPG, accelerometer, and temperature sensor, and is powered by a rechargeable battery. (b) Custom charging adapters shown with the watch and earring form factors [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) Duration of the dataset collected from each wearable device and the subset in which all four wearable devices are simultaneously available. (b) Distribution of ground-truth heart rate across the dataset. heart rate distribution in our dataset. To ensure heart rate label quality, the ECG windows are screened to reject segments with invalid or poor-quality ECG, usually caused by loose chest-strap contact… view at source ↗
Figure 4
Figure 4. Figure 4: Heart rate estimation results across four wearable devices, using the DCL model. As shown in [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Distribution of windows across accelerometer-derived motion-power bins and corresponding HR estimation MAE for the four device locations using the DCL model. 7 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Skin color type–wise comparison of heart-rate MAE across models for each device. We compute the average MAE across participants with different Fitzpatrick skin types [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
read the original abstract

Wearables are widely used for mobile health monitoring, and photoplethysmography (PPG) is a key sensing modality for heart rate and related physiological measurements. However, public in-the-wild PPG datasets remain largely wrist-centric or limited to short, controlled studies, constraining research on emerging wearable form factors. We present Multi-site PPG, an in-the-wild physiological dataset collected from four custom-developed unobtrusive wearables: a smart earring, ring, watch, and necklace. Each device records green and infrared reflective PPG, 3-axis acceleration, and temperature with timestamps for cross-device alignment, while a Polar H10 chest strap provides reference electrocardiogram (ECG). Participants wore the devices for multiple days during daytime activities while continuing their normal routines. The dataset contains over 350 hours of raw data and 230-290 hours of modeling-ready 8-second windows per wearable. We benchmark heuristic, supervised, and self-supervised heart-rate estimation methods, showing substantial body-site differences: the best methods achieve mean absolute errors (MAEs) of 2.30 bpm on the earring, 5.13 bpm on the ring, 8.37 bpm on the watch, and 8.68 bpm on the necklace. We further analyze motion effects and evaluate multi-site and PPG-accelerometer fusion, demonstrating the dataset's value for robust physiological sensing across emerging wearable form factors.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents the Multi-site PPG dataset, collected in-the-wild from four custom wearables (smart earring, ring, watch, necklace) that record green/infrared reflective PPG, 3-axis acceleration, and temperature, synchronized with Polar H10 ECG reference. Participants wore the devices during normal daytime routines for multiple days, yielding over 350 hours of raw data and 230-290 hours of 8-second modeling-ready windows per site. The authors benchmark heuristic, supervised, and self-supervised heart-rate estimation methods and report site-specific MAEs of 2.30 bpm (earring), 5.13 bpm (ring), 8.37 bpm (watch), and 8.68 bpm (necklace), along with analyses of motion effects and multi-site/PPG-accelerometer fusion.

Significance. If the reported MAE differences can be attributed to body-site properties rather than data-quality variations, the release of this large-scale public multi-site in-the-wild PPG dataset would be a meaningful contribution to mobile health and wearable sensing research. Most existing public PPG collections are wrist-centric or limited to short controlled sessions; this work addresses that gap by covering emerging form factors with raw signals, alignment timestamps, and pre-processed windows that support reproducibility. The concrete empirical benchmarks and motion/fusion analyses provide a useful starting point for developing robust methods across non-traditional sites.

major comments (2)
  1. [Results (MAE reporting and site comparisons)] Results section (MAE reporting and site comparisons): The central claim of substantial body-site differences rests on the MAEs (2.30 bpm earring to 8.68 bpm necklace). However, the manuscript provides no per-site statistics on PPG amplitude, fraction of windows rejected for poor contact, or motion-intensity distributions derived from the 3-axis accelerometer. Without these indicators, it remains possible that the performance gaps partly reflect differences in average skin contact quality or artifact severity across form factors rather than inherent optical or physiological site properties, especially under the free-living protocol.
  2. [Methods (benchmarking protocol)] Methods (benchmarking protocol): The evaluation of supervised and self-supervised models does not specify whether participant-level cross-validation was used or whether held-out windows were stratified by motion intensity. This detail is load-bearing for interpreting the generalizability of the reported MAEs and the validity of direct cross-site comparisons.
minor comments (2)
  1. [Abstract] Abstract: The phrase '230-290 hours of modeling-ready 8-second windows per wearable' would be more informative if broken down by individual site to allow immediate assessment of dataset balance.
  2. [Dataset description] Dataset description: Adding a summary table with total hours, participant count, and average windows per site would improve readability and quick reference.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. The comments highlight important aspects for strengthening the interpretation of site-specific differences and the evaluation protocol. We address each major comment below and outline the planned revisions.

read point-by-point responses

  1. Referee: [Results (MAE reporting and site comparisons)] Results section (MAE reporting and site comparisons): The central claim of substantial body-site differences rests on the MAEs (2.30 bpm earring to 8.68 bpm necklace). However, the manuscript provides no per-site statistics on PPG amplitude, fraction of windows rejected for poor contact, or motion-intensity distributions derived from the 3-axis accelerometer. Without these indicators, it remains possible that the performance gaps partly reflect differences in average skin contact quality or artifact severity across form factors rather than inherent optical or physiological site properties, especially under the free-living protocol.

    Authors: We agree that per-site statistics on PPG amplitude, rejection rates for poor contact, and motion-intensity distributions would help readers better attribute the observed MAE differences to body-site properties versus variations in signal quality or artifact levels. The released dataset contains the necessary raw signals to compute these metrics. In the revised manuscript, we will add a dedicated subsection and accompanying table in the Results section reporting: (1) mean and standard deviation of PPG amplitude per site, (2) the fraction of 8-second windows rejected due to poor contact (using the quality criteria described in the Methods), and (3) summary statistics and distributions of motion intensity derived from the 3-axis accelerometer. This addition will directly address the concern and support the central claims. revision: yes

  2. Referee: [Methods (benchmarking protocol)] Methods (benchmarking protocol): The evaluation of supervised and self-supervised models does not specify whether participant-level cross-validation was used or whether held-out windows were stratified by motion intensity. This detail is load-bearing for interpreting the generalizability of the reported MAEs and the validity of direct cross-site comparisons.

    Authors: We thank the referee for noting this omission. The benchmarking protocol employed participant-level 5-fold cross-validation, with all windows from any given participant assigned exclusively to either the training or test set. The folds were further stratified to preserve similar distributions of motion intensity (computed as the standard deviation of the acceleration magnitude) across training and test sets. We will revise the Methods section to explicitly document this protocol, including the number of folds, the participant-level split rule, and the stratification procedure based on motion intensity. This clarification will improve the interpretability of the reported MAEs and cross-site comparisons. revision: yes

Circularity Check

0 steps flagged

No significant circularity in empirical dataset benchmarks

full rationale

The paper releases a new multi-site PPG dataset collected from custom wearables during free-living activities and reports direct empirical MAEs from applying existing heuristic, supervised, and self-supervised heart-rate estimation methods to the collected signals. These performance numbers are measured outcomes on the new data rather than quantities derived from fitted parameters, self-citations, or any derivation chain that reduces to its own inputs by construction. No load-bearing steps invoke uniqueness theorems, ansatzes smuggled via prior work, or renamings of known results; the work is self-contained as a data release and benchmark against external reference ECG.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central contribution is an empirical dataset rather than a derivation; no free parameters are fitted to produce the headline MAEs, no new physical entities are postulated, and the only background assumptions are standard signal-processing practices for reflective PPG.

axioms (1)
  • domain assumption Reflective PPG at green and infrared wavelengths can be used to estimate heart rate when motion artifacts are present.
    Invoked implicitly when benchmarking heart-rate methods on the collected signals.

pith-pipeline@v0.9.0 · 5809 in / 1402 out tokens · 27631 ms · 2026-05-20T09:29:04.909476+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

56 extracted references · 56 canonical work pages

  1. [1]

    Apple Watch Series 11 - Technical Specifications

    Apple. Apple Watch Series 11 - Technical Specifications. https://www.apple.com/ apple-watch-series-11/specs/. Accessed: 2026-04-29

    work page 2026

  2. [2]

    Enhanceppg: Improving ppg-based heart rate estimation with self-supervision and augmentation

    Luca Benfenati, Sofia Belloni, Alessio Burrello, Panagiotis Kasnesis, Xiaying Wang, Luca Benini, Massimo Poncino, Enrico Macii, and Daniele Jahier Pagliari. Enhanceppg: Improving ppg-based heart rate estimation with self-supervision and augmentation. In2025 IEEE 7th International Conference on Artificial Intelligence Circuits and Systems (AICAS), pages 1–...

    work page 2025

  3. [3]

    Beliefppg: Uncertainty- aware heart rate estimation from ppg signals via belief propagation

    Valentin Bieri, Paul Streli, Berken Utku Demirel, and Christian Holz. Beliefppg: Uncertainty- aware heart rate estimation from ppg signals via belief propagation. InUncertainty in Artificial Intelligence, pages 173–183. PMLR, 2023

    work page 2023

  4. [4]

    Multi-scale peak and trough detection optimised for periodic and quasi-periodic neuroscience data

    SM Bishop and A Ercole. Multi-scale peak and trough detection optimised for periodic and quasi-periodic neuroscience data. 2018. doi: 10.17863/CAM.18051. URL https: //www.repository.cam.ac.uk/handle/1810/273187

    work page doi:10.17863/cam.18051 2018

  5. [5]

    Bishop and Ari Ercole

    Steven M. Bishop and Ari Ercole. Multi-scale peak and trough detection optimised for peri- odic and quasi-periodic neuroscience data. InIntracranial Pressure & Neuromonitoring XVI. Springer, 2018. doi: 10.1007/978-3-319-65798-1_39

    work page doi:10.1007/978-3-319-65798-1_39 2018

  6. [6]

    Dwaipayan Biswas, Luke Everson, Muqing Liu, Madhuri Panwar, Bram-Ernst Verhoef, Shrishail Patki, Chris H Kim, Amit Acharyya, Chris Van Hoof, Mario Konijnenburg, et al. Cornet: Deep learning framework for ppg-based heart rate estimation and biometric identification in ambulant environment.IEEE transactions on biomedical circuits and systems, 13(2):282–291, 2019

    work page 2019

  7. [7]

    The fourier transform.Scientific American, 260(6):86–95, 1989

    Ronald N Bracewell. The fourier transform.Scientific American, 260(6):86–95, 1989

    work page 1989

  8. [8]

    Rui Cao, Iman Azimi, Fatemeh Sarhaddi, Hannakaisa Niela-Vilen, Anna Axelin, Pasi Liljeberg, and Amir M Rahmani. Accuracy assessment of oura ring nocturnal heart rate and heart rate vari- ability in comparison with electrocardiography in time and frequency domains: comprehensive analysis.Journal of Medical Internet Research, 24(1):e27487, 2022

    work page 2022

  9. [9]

    Detecting beats in the photoplethysmogram: benchmarking open-source algorithms.Physiological Measurement, 43 (8):085007, 2022

    Peter H Charlton, Kevin Kotzen, Elisa Mejía-Mejía, Philip J Aston, Karthik Budidha, Jonathan Mant, Callum Pettit, Joachim A Behar, and Panicos A Kyriacou. Detecting beats in the photoplethysmogram: benchmarking open-source algorithms.Physiological Measurement, 43 (8):085007, 2022

    work page 2022

  10. [10]

    Deep learning for heart rate estimation from reflectance photoplethysmography with acceleration power spectrum and acceleration intensity.Ieee Access, 8:63390–63402, 2020

    Heewon Chung, Hoon Ko, Hooseok Lee, and Jinseok Lee. Deep learning for heart rate estimation from reflectance photoplethysmography with acceleration power spectrum and acceleration intensity.Ieee Access, 8:63390–63402, 2020

    work page 2020

  11. [11]

    Siamquality: a convnet-based foundation model for photoplethysmography signals.Physiologi- cal Measurement, 45(8):085004, 2024

    Cheng Ding, Zhicheng Guo, Zhaoliang Chen, Randall J Lee, Cynthia Rudin, and Xiao Hu. Siamquality: a convnet-based foundation model for photoplethysmography signals.Physiologi- cal Measurement, 45(8):085004, 2024

    work page 2024

  12. [12]

    Systolic peak detection in acceleration photoplethysmograms measured from emergency responders in tropical conditions.PLOS ONE, 8(10):e76585, 2013

    Mohamed Elgendi, Ian Norton, Matt Brearley, Derek Abbott, and Chris Schacht. Systolic peak detection in acceleration photoplethysmograms measured from emergency responders in tropical conditions.PLOS ONE, 8(10):e76585, 2013. doi: 10.1371/journal.pone.0076585

    work page doi:10.1371/journal.pone.0076585 2013

  13. [13]

    Wf-ppg: A wrist-finger dual-channel dataset for studying the impact of contact pressure on ppg morphology.Scientific Data, 12(1):200, 2025

    Matthew Yiwen Ho, Hung Manh Pham, Aaqib Saeed, and Dong Ma. Wf-ppg: A wrist-finger dual-channel dataset for studying the impact of contact pressure on ppg morphology.Scientific Data, 12(1):200, 2025

    work page 2025

  14. [14]

    Heart rate estimation in ppg signals using convolutional-recurrent regressor.Computers in Biology and Medicine, 145:105470, 2022

    Shahid Ismail, Imran Siddiqi, and Usman Akram. Heart rate estimation in ppg signals using convolutional-recurrent regressor.Computers in Biology and Medicine, 145:105470, 2022

    work page 2022

  15. [15]

    Description of a database containing wrist ppg signals recorded during physical exercise with both accelerometer and gyroscope measures of motion

    Delaram Jarchi and Alexander J Casson. Description of a database containing wrist ppg signals recorded during physical exercise with both accelerometer and gyroscope measures of motion. Data, 2(1):1, 2016. 10

    work page 2016

  16. [16]

    Multi-head cross-attentional ppg and motion signal fusion for heart rate estimation.arXiv preprint arXiv:2210.11415, 2022

    Panagiotis Kasnesis, Lazaros Toumanidis, Alessio Burrello, Christos Chatzigeorgiou, and Charalampos Z Patrikakis. Multi-head cross-attentional ppg and motion signal fusion for heart rate estimation.arXiv preprint arXiv:2210.11415, 2022

    work page arXiv 2022

  17. [17]

    Feature-level cross-attentional ppg and motion signal fusion for heart rate estimation

    Panagiotis Kasnesis, Lazaros Toumanidis, Alessio Burrello, Christos Chatzigeorgiou, and Charalampos Z Patrikakis. Feature-level cross-attentional ppg and motion signal fusion for heart rate estimation. In2023 IEEE 47th Annual Computers, Software, and Applications Conference (COMPSAC), pages 1731–1736. IEEE, 2023

    work page 2023

  18. [18]

    Daniel Koerber, Shawn Khan, Tahmina Shamsheri, Abirami Kirubarajan, and Sangeeta Mehta. Accuracy of heart rate measurement with wrist-worn wearable devices in various skin tones: a systematic review.Journal of Racial and Ethnic Health Disparities, 10(6):2676–2684, 2023

    work page 2023

  19. [19]

    Motion artifact cancellation in wearable photoplethysmography using gyroscope.IEEE Sensors Journal, 19(3):1166–1175, 2018

    H Lee, H Chung, and J Lee. Motion artifact cancellation in wearable photoplethysmography using gyroscope.IEEE Sensors Journal, 19(3):1166–1175, 2018

    work page 2018

  20. [20]

    B. N. Li, M. C. Dong, and M. I. Vai. On an automatic delineator for arterial blood pressure waveforms.Biomedical Signal Processing and Control, 5(1):76–81, 2010. doi: 10.1016/j.bspc. 2009.06.002

    work page doi:10.1016/j.bspc 2010

  21. [21]

    Lau, Jan C

    Dominique Makowski, Tam Pham, Zen J. Lau, Jan C. Brammer, François Lespinasse, Hung Pham, Christopher Schölzel, and S. H. Annabel Chen. Neurokit2: A python toolbox for neurophysiological signal processing.Behavior Research Methods, 53(4):1689–1696, 2021. doi: 10.3758/s13428-020-01516-y

    work page doi:10.3758/s13428-020-01516-y 2021

  22. [22]

    Scamps: Synthetics for camera measurement of physiological signals

    Daniel McDuff, Miah Wander, Xin Liu, Brian Hill, Javier Hernandez, Jonathan Lester, and Tadas Baltrusaitis. Scamps: Synthetics for camera measurement of physiological signals. Advances in Neural Information Processing Systems, 35:3744–3757, 2022

    work page 2022

  23. [23]

    Milad Asgari Mehrabadi, Iman Azimi, Fatemeh Sarhaddi, Anna Axelin, Hannakaisa Niela- Vilén, Saana Myllyntausta, Sari Stenholm, Nikil Dutt, Pasi Liljeberg, and Amir M Rahmani. Sleep tracking of a commercially available smart ring and smartwatch against medical-grade actigraphy in everyday settings: instrument validation study.JMIR mHealth and uHealth, 8 (1...

    work page 2020

  24. [24]

    Manuel Meier and Christian Holz. Assessing the accuracy of photoplethysmography for wearable heart rate monitoring based on body location and body motion in uncontrolled outdoor environments.Current Issues in Sport Science (CISS), 9(2):059–059, 2024

    work page 2024

  25. [25]

    Robust heart rate detection via multi-site photoplethysmog- raphy

    Manuel Meier and Christian Holz. Robust heart rate detection via multi-site photoplethysmog- raphy. In2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 1–4. IEEE, 2024

    work page 2024

  26. [26]

    Wildppg: a real-world ppg dataset of long continuous recordings.Advances in Neural Information Processing Systems, 37:2246– 2266, 2024

    Manuel Meier, Berken Utku Demirel, and Christian Holz. Wildppg: a real-world ppg dataset of long continuous recordings.Advances in Neural Information Processing Systems, 37:2246– 2266, 2024

    work page 2024

  27. [27]

    Earset: A multi-modal dataset for studying the impact of head and facial movements on in-ear ppg signals.Scientific data, 10(1):850, 2023

    Alessandro Montanari, Andrea Ferlini, Ananta Narayanan Balaji, Cecilia Mascolo, and Fahim Kawsar. Earset: A multi-modal dataset for studying the impact of head and facial movements on in-ear ppg signals.Scientific data, 10(1):850, 2023

    work page 2023

  28. [28]

    Scaling wearable foundation models

    Girish Narayanswamy, Xin Liu, Kumar Ayush, Yuzhe Yang, Xuhai Xu, Jake Garrison, Shyam A Tailor, Jacob Sunshine, Yun Liu, Tim Althoff, et al. Scaling wearable foundation models. In The Thirteenth International Conference on Learning Representations, 2025

    work page 2025

  29. [29]

    Andrea Nemcova, Enikö Vargova, Radovan Smisek, Lucie Marsanova, Lukas Smital, and Martin Vitek. Brno university of technology smartphone ppg database (but ppg): Annotated dataset for ppg quality assessment and heart rate estimation.BioMed Research International, 2021(1):3453007, 2021

    work page 2021

  30. [30]

    Brno University of Technology Smartphone PPG Database (BUT PPG).PhysioNet, August 2024

    Andrea Nemcova, Radovan Smisek, Eniko Vargova, Lucie Maršánová, Martin Vitek, Lukas Smital, Marina Filipenska, Pavlina Sikorova, and Pavel Gálík. Brno University of Technology Smartphone PPG Database (BUT PPG).PhysioNet, August 2024. doi: 10.13026/tn53-8153. URLhttps://doi.org/10.13026/tn53-8153. Version 2.0.0. 11

    work page doi:10.13026/tn53-8153 2024

  31. [31]

    Wavelet analysis and self-similarity of photoplethysmography signals for hrv estimation and quality assessment.Sensors, 21(20):6798, 2021

    Alexander Neshitov, Konstantin Tyapochkin, Evgeniya Smorodnikova, and Pavel Pravdin. Wavelet analysis and self-similarity of photoplethysmography signals for hrv estimation and quality assessment.Sensors, 21(20):6798, 2021

    work page 2021

  32. [32]

    Ppg-distill: Efficient photoplethysmography signals analysis via foundation model distillation.arXiv preprint arXiv:2509.19215, 2025

    Juntong Ni, Saurabh Kataria, Shengpu Tang, Carl Yang, Xiao Hu, and Wei Jin. Ppg-distill: Efficient photoplethysmography signals analysis via foundation model distillation.arXiv preprint arXiv:2509.19215, 2025

    work page arXiv 2025

  33. [33]

    Oura Ring 4

    Oura. Oura Ring 4. https://support.ouraring.com/hc/en-us/articles/ 33045011508115-Oura-Ring-4. Accessed: 2026-04-29

    work page 2026

  34. [34]

    Oura Ring: Smart Ring for Fitness, Stress, Sleep, and Health.https://ouraring.com/,

    Oura. Oura Ring: Smart Ring for Fitness, Stress, Sleep, and Health.https://ouraring.com/,

    work page

  35. [35]

    Official website, accessed April 2, 2026

    work page 2026

  36. [36]

    Tompkins

    Jiapu Pan and Willis J. Tompkins. A real-time qrs detection algorithm.IEEE Transactions on Biomedical Engineering, BME-32(3):230–236, 1985. doi: 10.1109/TBME.1985.325532

    work page doi:10.1109/tbme.1985.325532 1985

  37. [37]

    A ppg signal dataset collected in semi- naturalistic settings using galaxy watch.Scientific Data, 12(1):892, 2025

    Sangjun Park, Dejiang Zheng, and Uichin Lee. A ppg signal dataset collected in semi- naturalistic settings using galaxy watch.Scientific Data, 12(1):892, 2025. doi: 10.1038/ s41597-025-05152-z

    work page 2025

  38. [38]

    A ppg signal dataset collected in semi-naturalistic settings using galaxy watch.Scientific Data, 12(1):892, 2025

    Sangjun Park, Dejiang Zheng, and Uichin Lee. A ppg signal dataset collected in semi-naturalistic settings using galaxy watch.Scientific Data, 12(1):892, 2025

    work page 2025

  39. [39]

    Toward a robust estimation of respiratory rate from pulse oximeters.IEEE Transactions on Biomedical Engineering, 64(8):1914–1923, 2016

    Marco AF Pimentel, Alistair EW Johnson, Peter H Charlton, Drew Birrenkott, Peter J Watkinson, Lionel Tarassenko, and David A Clifton. Toward a robust estimation of respiratory rate from pulse oximeters.IEEE Transactions on Biomedical Engineering, 64(8):1914–1923, 2016

    work page 1914

  40. [40]

    PPG-DaLiA

    Attila Reiss, Ina Indlekofer, and Philip Schmidt. PPG-DaLiA. UCI Machine Learning Reposi- tory, 2019. DOI: https://doi.org/10.24432/C53890

    work page doi:10.24432/c53890 2019

  41. [41]

    Deep ppg: Large-scale heart rate estimation with convolutional neural networks.Sensors, 19(14):3079, 2019

    Attila Reiss, Ina Indlekofer, Philip Schmidt, and Kristof Van Laerhoven. Deep ppg: Large-scale heart rate estimation with convolutional neural networks.Sensors, 19(14):3079, 2019

    work page 2019

  42. [42]

    Mithun Saha, Maxwell A Xu, Wanting Mao, Sameer Neupane, James M Rehg, and Santosh Kumar. Pulse-ppg: An open-source field-trained ppg foundation model for wearable applications across lab and field settings.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 9(3):1–35, 2025

    work page 2025

  43. [43]

    Introducing wesad, a multimodal dataset for wearable stress and affect detection

    Philip Schmidt, Attila Reiss, Robert Duerichen, Claus Marberger, and Kristof Van Laerhoven. Introducing wesad, a multimodal dataset for wearable stress and affect detection. InProceedings of the 20th ACM international conference on multimodal interaction, pages 400–408, 2018

    work page 2018

  44. [44]

    A dataset and toolkit for multiparameter cardiovascular physiology sensing on rings.arXiv preprint arXiv:2505.04172, 2025

    Jiankai Tang, Kegang Wang, Yingke Ding, Jiatong Ji, Zeyu Wang, Xiyuxing Zhang, Ping Chen, Yuanchun Shi, and Yuntao Wang. A dataset and toolkit for multiparameter cardiovascular physiology sensing on rings.arXiv preprint arXiv:2505.04172, 2025

    work page arXiv 2025

  45. [45]

    Heartpy: A novel heart rate algorithm for the analysis of noisy signals.Transportation research part F: traffic psychology and behaviour, 66:368–378, 2019

    Paul Van Gent, Haneen Farah, Nicole Van Nes, and Bart Van Arem. Heartpy: A novel heart rate algorithm for the analysis of noisy signals.Transportation research part F: traffic psychology and behaviour, 66:368–378, 2019

    work page 2019

  46. [46]

    Heartpy: A novel heart rate algorithm for the analysis of noisy signals.Transportation Research Part F: Traffic Psychology and Behaviour, 66:368–378, 2019

    Paul van Gent, Haneen Farah, Nicole van Nes, and Bart van Arem. Heartpy: A novel heart rate algorithm for the analysis of noisy signals.Transportation Research Part F: Traffic Psychology and Behaviour, 66:368–378, 2019. doi: 10.1016/j.trf.2019.09.015

    work page doi:10.1016/j.trf.2019.09.015 2019

  47. [47]

    An open source benchmarked toolbox for cardiovascular waveform and interval analysis.Physiological measurement, 39(10):105004, 2018

    Adriana N Vest, Giulia Da Poian, Qiao Li, Chengyu Liu, Shamim Nemati, Amit J Shah, and Gari D Clifford. An open source benchmarked toolbox for cardiovascular waveform and interval analysis.Physiological measurement, 39(10):105004, 2018

    work page 2018

  48. [48]

    Vest, Giulia Da Poian, Qiao Li, Chengyu Liu, Shamim Nemati, A

    Andrew N. Vest, Giulia Da Poian, Qiao Li, Chengyu Liu, Shamim Nemati, A. J. Shah, and Gari D. Clifford. An open source benchmarked toolbox for cardiovascular waveform and interval analysis.Physiological Measurement, 39(10):105004, 2018. doi: 10.1088/1361-6579/aae021. 12

    work page doi:10.1088/1361-6579/aae021 2018

  49. [49]

    Time series classification from scratch with deep neural networks: A strong baseline

    Zhiguang Wang, Weizhong Yan, and Tim Oates. Time series classification from scratch with deep neural networks: A strong baseline. In2017 International joint conference on neural networks (IJCNN), pages 1578–1585. IEEE, 2017

    work page 2017

  50. [50]

    Multi-headed conv-lstm network for heart rate estimation during daily living activities.Sensors, 21(15):5212, 2021

    Michał Wilkosz and Agnieszka Szcz˛ esna. Multi-headed conv-lstm network for heart rate estimation during daily living activities.Sensors, 21(15):5212, 2021

    work page 2021

  51. [51]

    Ppg earring: Wireless smart earring for heart health monitoring

    Qiuyue Xue, Dilini Nissanka, Jiachen Tammy Yan, Ruiqing Wang, Shwetak Patel, and Vikram Iyer. Ppg earring: Wireless smart earring for heart health monitoring. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pages 1–16, 2025

    work page 2025

  52. [52]

    Unsupervised represen- tation learning on high-dimensional clinical data improves genomic discovery and prediction

    Taedong Yun, Justin Cosentino, Babak Behsaz, Zachary R McCaw, Davin Hill, Robert Luben, Dongbing Lai, John Bates, Howard Yang, Tae-Hwi Schwantes-An, et al. Unsupervised represen- tation learning on high-dimensional clinical data improves genomic discovery and prediction. Nature Genetics, 56(8):1604–1613, 2024

    work page 2024

  53. [53]

    Zexing Zhang, Huimin Lu, Songzhe Ma, Jianzhong Peng, Chenglin Lin, Niya Li, and Bingwang Dong. A general framework for generative self-supervised learning in non-invasive estimation of physiological parameters using photoplethysmography.Biomedical Signal Processing and Control, 98:106788, 2024

    work page 2024

  54. [54]

    Photoplethysmography-based heart rate monitoring in physical activities via joint sparse spectrum reconstruction.IEEE transactions on biomedical engineering, 62(8): 1902–1910, 2015

    Zhilin Zhang. Photoplethysmography-based heart rate monitoring in physical activities via joint sparse spectrum reconstruction.IEEE transactions on biomedical engineering, 62(8): 1902–1910, 2015

    work page 1902

  55. [55]

    Troika: A general framework for heart rate monitoring using wrist-type photoplethysmographic signals during intensive physical exercise

    Zhilin Zhang, Zhouyue Pi, and Benyuan Liu. Troika: A general framework for heart rate monitoring using wrist-type photoplethysmographic signals during intensive physical exercise. IEEE Transactions on biomedical engineering, 62(2):522–531, 2014

    work page 2014

  56. [56]

    Signal quality assessment of electrocardiograms: a compre- hensive review and empirical investigation.Journal of Electrocardiology, 51(6):S91–S97, 2018

    Zhidong Zhao and Yilong Zhang. Signal quality assessment of electrocardiograms: a compre- hensive review and empirical investigation.Journal of Electrocardiology, 51(6):S91–S97, 2018. doi: 10.1016/j.jelectrocard.2018.08.004. 13 Appendix A Dataset Accessibility We host the dataset on Hugging Face in .npz format and provide Python loaders, preprocessing scr...

    work page doi:10.1016/j.jelectrocard.2018.08.004 2018