Paperdoi:10.17632/fhp8brnycd.1
Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap
Mendeley Data companion record (manuscript not yet located publicly)
TL;DR
Pre-registered negative result on wrist-PPG heart rate: across four public corpora, swapping the R-peak detector or the HR aggregation formula shifts the reference label by under 0.5 bpm, while zero-shot cross-dataset error averages 18.65 bpm. A frozen PaPaGei encoder with a ridge probe does no better than FFT features across datasets (20.3 vs 19.2 bpm).
- Form factor: wrist
Abstract
Companion research data for a paired dual-label benchmark of cross-dataset wrist-PPG heart-rate estimation. Contains two pre-registered frozen-encoder protocols, the analysis and independent gate scripts, and aggregated result tables and machine-readable verdicts. No raw physiological recordings, no window-level feature tables and no third-party dataset copies are included; the four source corpora must be obtained from their original repositories (DOIs cited in the manuscript). Scripts use a placeholder for the local project directory.
Verbatim from DataCite (opens in a new tab) · Licence: info:eu-repo/semantics/openAccess
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Devices
Keywords
Datasets used
IEEE Signal Processing Cup 2015 (TROIKA)wrist PPG during intensive exerciseTrainingEvaluation12 subjects · PPG · ECG · IMU
PPG-DaLiAPPG dataset for motion compensation / HRTrainingEvaluation15 subjects · PPG · ECG · IMU
WESADWearable Stress and Affect DetectionTrainingEvaluation15 subjects · PPG · ECG · IMU
Wrist PPG During Exercise (PhysioNet wrist)TrainingEvaluation8 subjects · PPG · ECG · IMU
How it works
Cite this paper
Wang, S., Zhang, J., & Wang, Y. (2026). Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap. Mendeley Data companion record. https://doi.org/10.17632/fhp8brnycd.1
Wang, Songfeng, Jinling Zhang, and Yuhua Wang. 2026. "Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap." Mendeley Data companion record. https://doi.org/10.17632/fhp8brnycd.1.
Wang, S., Zhang, J. and Wang, Y. (2026) 'Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap', Mendeley Data companion record. doi: 10.17632/fhp8brnycd.1.
Wang, Songfeng, et al. "Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap." Mendeley Data companion record, 2026. https://doi.org/10.17632/fhp8brnycd.1.
S. Wang, J. Zhang, and Y. Wang, "Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap," Mendeley Data companion record, 2026, doi: 10.17632/fhp8brnycd.1.
@article{wang2026analysis,
title = {{Analysis code, pre-registered protocols and aggregated result tables for "A Paired Dual-Label Benchmark Shows That Label Provenance Cannot Explain the Cross-Dataset PPG Heart-Rate Gap}},
author = {Wang, Songfeng and Zhang, Jinling and Wang, Yuhua},
year = {2026},
journal = {Mendeley Data companion record},
doi = {10.17632/fhp8brnycd.1}
}Added to the ChatPPG research hub