A Comprehensive Inference-Time Augmentation Framework in Physiological Signals: Application to PPG-Based AF Detection

Davood Fattahi; Runze Yan; Saurabh Kataria; Xiao Hu; Zhaoliang Chen

arxiv: 2606.10410 · v1 · pith:E3UPGFWLnew · submitted 2026-06-09 · 💻 cs.LG · eess.SP· q-bio.QM

A Comprehensive Inference-Time Augmentation Framework in Physiological Signals: Application to PPG-Based AF Detection

Davood Fattahi , Runze Yan , Saurabh Kataria , Zhaoliang Chen , Xiao Hu This is my paper

Pith reviewed 2026-06-27 14:06 UTC · model grok-4.3

classification 💻 cs.LG eess.SPq-bio.QM

keywords inference-time augmentationPPGatrial fibrillation detectionphysiological signalsrobustnessBayesian optimizationmachine learning

0 comments

The pith

A framework of 13 Bayesian-optimized augmentations applied at inference improves robustness in PPG-based atrial fibrillation detection without retraining.

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

Inference-time augmentation applies data transformations during prediction rather than training to counter noise, artifacts, and distribution shifts in physiological signals. The paper presents a unified framework that combines 13 methods from time, amplitude, frequency, and artifact domains, tunes their hyperparameters through Bayesian optimization, and tests the result on atrial fibrillation classification from 30-second PPG recordings. Across GPT-PPG and ResNet models and five datasets covering more than 400 patients, standard ITA raised AUROC by as much as 8.5 percent and AUPRC by 10.6 percent, while selective ITA lowered false-positive rates on non-AF recordings. A reader would care because many clinical systems cannot retrain models when new sensors or patient groups appear, so an inference-only adjustment offers a direct route to greater reliability.

Core claim

The paper establishes that a comprehensive inference-time augmentation framework, consisting of 13 augmentation techniques with Bayesian-optimized hyperparameters, improves the performance of PPG-based atrial fibrillation detection. Standard application of the augmentations during inference increases area under the ROC curve by up to 8.5% for GPT-PPG and 0.7% for ResNet, and area under the precision-recall curve by up to 10.6% and 0.8% respectively. Selective use of the augmentations further decreases average false positive rates by up to 4.4% and 1.3% on non-AF datasets. These gains demonstrate that ITA provides a model-agnostic way to enhance reliability in settings where retraining is imp

What carries the argument

The unified inference-time augmentation framework that applies 13 methods spanning time-domain, amplitude-domain, frequency-domain, and artifact-injection transformations, with hyperparameters selected by Bayesian optimization.

If this is right

Standard ITA raises AUROC and AUPRC on the AF detection task for both tested architectures.
Selective ITA lowers average false-positive rates on recordings that do not contain atrial fibrillation.
The same augmentation set works without modification for GPT-PPG and ResNet.
The reported gains hold across five independent datasets that together exceed 400 patients and 9,800 hours of recording.
All improvements occur at inference time and require no change to model weights or training procedure.

Where Pith is reading between the lines

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

The same set of augmentations could be applied to other physiological modalities such as ECG or respiratory signals where motion artifacts are common.
Clinical teams facing new hardware or demographic shifts could adopt ITA to extend the usable life of existing trained models rather than collecting fresh labeled data.
Evaluating the framework on streaming, variable-length recordings instead of fixed 30-second windows would test whether the reported gains persist under continuous monitoring conditions.

Load-bearing premise

The performance gains from the 13 augmentation methods and their Bayesian-optimized hyperparameters will continue to appear on real-world PPG data drawn from distributions outside the five evaluation datasets.

What would settle it

Running the same models and selective ITA procedure on a new PPG dataset collected with a different sensor type or patient population and observing no statistically significant rise in AUROC or drop in false-positive rate relative to the unaugmented baseline.

read the original abstract

Objective: Accurate classification of physiological signals in real-world deployments is challenged by sensor noise, motion artifacts, and distribution shifts between training and deployment data. Inference-time augmentation (ITA), which applies augmentations during inference rather than retraining, offers a simple, model-agnostic mechanism to improve robustness. However, ITA application to physiological signals has remained narrow in scope, relying on limited augmentation methods with fixed, unoptimized parameters. This work proposes a unified ITA framework to address that gap. Approach: The framework incorporates 13 augmentation methods spanning time-domain, amplitude-domain, frequency-domain, and artifact-injection transformations, with hyperparameters optimized via Bayesian optimization. We evaluate on atrial fibrillation (AF) detection from 30-second PPG signals using GPT-PPG and ResNet across five datasets comprising more than 400 patients and ${\sim}$9,800 hours of recording. Main results: Standard ITA consistently improved AUROC (up to 8.5% for GPT-PPG and 0.7% for ResNet) and AUPRC (up to 10.6% for GPT-PPG and 0.8% for ResNet). Selective ITA further reduced average FPR by up to 4.4% (GPT-PPG) and 1.3% (ResNet) on non-AF datasets. Significance: These findings establish ITA as a practical, model-agnostic approach for improving PPG-based AF classification reliability in deployment settings where retraining is not feasible, with broader applicability to physiological signal analysis.

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.

Desk Editor's Note private letter to a colleague

Bayesian-tuned ITA across 13 methods lifts PPG AF detection numbers on GPT-PPG but the optimization protocol and model dependence are the parts to watch.

read the letter

The main thing here is that a broad set of inference-time augmentations, tuned by Bayesian search, produces measurable gains on atrial fibrillation detection from 30-second PPG segments, with the biggest lifts on the transformer model.

The new piece is the scope. Prior ITA work on physiological signals used only a few fixed transforms. This paper assembles 13 methods across time, amplitude, frequency, and artifact domains and optimizes their parameters. They also add a selective mode that skips augmentation when it would increase false positives on non-AF recordings. The evaluation runs on five datasets covering more than 400 patients and roughly 9,800 hours, using both GPT-PPG and ResNet. That setup is more systematic than earlier narrow applications.

The reported improvements are consistent in the abstract: up to 8.5% AUROC and 10.6% AUPRC on GPT-PPG, smaller but still positive on ResNet, plus FPR drops from the selective version. The multi-dataset, multi-model design gives the numbers some weight.

The soft spot is the hyperparameter search. Bayesian optimization over 13 methods can pick up dataset-specific noise or demographics if the search is not strictly limited to training and validation splits. The abstract does not state the exact protocol, so it is unclear whether test-set statistics influenced the chosen parameters. If they did, the claimed robustness for unseen deployment data would be overstated. The gains also shrink sharply on ResNet, which suggests the framework is not uniformly helpful across architectures.

This is for engineers who deploy physiological classifiers and cannot retrain on new hardware or patient cohorts. Readers who want concrete, model-agnostic robustness tricks will find the framework and the numbers useful. It deserves a serious referee because the empirical core is straightforward and the experiments span multiple conditions.

Send it to review, but request explicit confirmation that the Bayesian search stayed inside train/validation splits and that the hyperparameters were not tuned with any test data access.

Referee Report

1 major / 2 minor

Summary. The manuscript proposes a unified inference-time augmentation (ITA) framework for physiological signals, applied to PPG-based atrial fibrillation detection. It incorporates 13 augmentation methods spanning time-, amplitude-, frequency-, and artifact-domains, with hyperparameters optimized via Bayesian optimization. Evaluations using GPT-PPG and ResNet models across five datasets (>400 patients, ~9,800 hours) report consistent AUROC gains (up to 8.5% for GPT-PPG, 0.7% for ResNet), AUPRC gains (up to 10.6% and 0.8%), and further FPR reductions via selective ITA on non-AF data.

Significance. If the results hold after addressing optimization details, the work establishes ITA as a practical, model-agnostic tool for improving robustness to noise and distribution shifts in physiological signal classification without retraining. The multi-dataset, multi-model evaluation and distinction between standard and selective ITA provide concrete evidence of utility in deployment scenarios, with potential extension to other signal types.

major comments (1)

[Approach] Approach section: The description of Bayesian optimization for the 13 augmentation hyperparameters does not explicitly state that the search was confined to training and validation splits strictly separated from the test portions of each dataset. This detail is load-bearing for the central claim of generalizable gains, as access to test-set statistics could allow selection of parameters tuned to the specific noise profiles or demographics in the evaluated recordings.

minor comments (2)

[Results] Results section: The reported performance improvements would be strengthened by inclusion of statistical significance tests (e.g., paired t-tests or Wilcoxon tests across folds/datasets) and error bars or confidence intervals on the AUROC/AUPRC deltas.
The manuscript would benefit from a table summarizing the 13 augmentation methods, their domains, and the final optimized hyperparameter values for reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for this constructive comment on the optimization procedure. We address the point below and confirm that the manuscript will be revised for clarity.

read point-by-point responses

Referee: The description of Bayesian optimization for the 13 augmentation hyperparameters does not explicitly state that the search was confined to training and validation splits strictly separated from the test portions of each dataset. This detail is load-bearing for the central claim of generalizable gains, as access to test-set statistics could allow selection of parameters tuned to the specific noise profiles or demographics in the evaluated recordings.

Authors: We confirm that Bayesian optimization was performed exclusively within the training and validation portions of each dataset using internal cross-validation, with no access to any test data or test-set statistics. This ensures the reported gains reflect generalization rather than overfitting to test distributions. We will revise the Approach section to explicitly document this data separation, including the cross-validation scheme used for hyperparameter search. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical evaluation on held-out data is self-contained.

full rationale

The paper proposes an empirical ITA framework consisting of 13 augmentation methods whose hyperparameters are tuned via Bayesian optimization and then applied at inference time. Reported gains in AUROC, AUPRC, and FPR are measured on held-out portions of five datasets rather than being quantities defined by construction from the fitted parameters themselves. No derivation chain, uniqueness theorem, self-citation load-bearing premise, or renaming of known results appears in the abstract or described approach; the central claims rest on standard experimental comparison that remains independent of the target performance metrics.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on empirical selection and tuning of augmentations rather than first-principles derivation; the key unverified premise is that the chosen transformations adequately cover real deployment variations.

free parameters (1)

Hyperparameters for each of the 13 augmentation methods
Tuned via Bayesian optimization on validation data to maximize performance metrics.

axioms (1)

domain assumption The 13 augmentation methods represent plausible real-world sensor noise, motion artifacts, and distribution shifts in PPG signals.
Invoked to justify why ITA improves robustness in deployment settings.

pith-pipeline@v0.9.1-grok · 5831 in / 1342 out tokens · 24000 ms · 2026-06-27T14:06:49.498784+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

59 extracted references · 44 canonical work pages

[1]

Understanding Test-Time Augmentation,

M. Kimura, "Understanding Test-Time Augmentation," in Neural Information Processing, Cham, T. Mantoro, M. Lee, M. A. Ayu, K. W. Wong, and A. N. Hidayanto, Eds., 2021// 2021: Springer International Publishing, pp. 558-569

2021
[2]

A survey on Image Data Augmentation for Deep Learning,

C. Shorten and T. M. Khoshgoftaar, "A survey on Image Data Augmentation for Deep Learning," Journal of Big Data, vol. 6, no. 1, p. 60, 2019/07/06 2019, doi: 10.1186/s40537-019-0197- 0

work page doi:10.1186/s40537-019-0197- 2019
[3]

Data augmentation of wearable sensor data for parkinson’s disease monitoring using convolutional neural networks,

T. T. Um et al., "Data augmentation of wearable sensor data for parkinson’s disease monitoring using convolutional neural networks," presented at the Proceedings of the 19th ACM International Conference on Multimodal Interaction, Glasgow, UK, 2017. [Online]. Available: https://doi.org/10.1145/3136755.3136817

work page doi:10.1145/3136755.3136817 2017
[4]

A Multimodal-Driven Fusion Data Augmentation Framework for Emotion Recognition,

A. Li, M. Wu, R. Ouyang, Y . Wang, F. Li, and Z. Lv, "A Multimodal-Driven Fusion Data Augmentation Framework for Emotion Recognition," IEEE Transactions on Artificial Intelligence, vol. 6, no. 8, pp. 2083-2097, 2025, doi: 10.1109/TAI.2025.3537965

work page doi:10.1109/tai.2025.3537965 2083
[5]

A novel data augmentation strategy for robust deep learning classification of biomedical time-series data: Application to ECG and EEG analysis,

M. Guhdar, R. J. Mstafa, and A. O. Mohammed, "A novel data augmentation strategy for robust deep learning classification of biomedical time-series data: Application to ECG and EEG analysis," arXiv [eess.SP], 2025/7/16 2025. [Online]. Available: http://arxiv.org/abs/2507.12645

Pith/arXiv arXiv 2025
[6]

In: Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, volume 2022-July

J. An, R. E. Gregg, and S. Borhani, "Effective Data Augmentation, Filters, and Automation Techniques for Automatic 12-Lead ECG Classification Using Deep Residual Neural Networks," in 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 11-15 July 2022 2022, pp. 1283-1287, doi: 10.1109/EMBC48229.2022.9871654

work page doi:10.1109/embc48229.2022.9871654 2022
[7]

Explanations of Augmentation Methods for Deep Learning ECG Classification,

N. S. P. Balasubramanian and S. Dakshit, "Explanations of Augmentation Methods for Deep Learning ECG Classification," in Artificial Intelligence in Medicine, Cham, J. Finkelstein, R. Moskovitch, and E. Parimbelli, Eds., 2024// 2024: Springer Nature Switzerland, pp. 277-287

2024
[8]

A novel data augmentation method to enhance deep neural networks for detection of atrial fibrillation,

P. Cao et al., "A novel data augmentation method to enhance deep neural networks for detection of atrial fibrillation," Biomedical Signal Processing and Control, vol. 56, p. 101675, 2020/02/01/ 2020, doi: https://doi.org/10.1016/j.bspc.2019.101675

work page doi:10.1016/j.bspc.2019.101675 2020
[9]

Emotion Detection Using Physiological Signals,

A. Dessai and H. Virani, "Emotion Detection Using Physiological Signals," in 2021 International Conference on Electrical, Computer and Energy Technologies (ICECET), 9- 10 Dec. 2021 2021, pp. 1-4, doi: 10.1109/ICECET52533.2021.9698729. Journal XX (XXXX) XXXXXX Author et al 17

work page doi:10.1109/icecet52533.2021.9698729 2021
[10]

Data Augmentation for 12-Lead ECG Beat Classification,

E. Do, J. Boynton, B. S. Lee, and D. Lustgarten, "Data Augmentation for 12-Lead ECG Beat Classification," SN Computer Science, vol. 3, no. 1, p. 70, 2021/11/19 2021, doi: 10.1007/s42979-021-00924-x

work page doi:10.1007/s42979-021-00924-x 2021
[11]

Empirical Study of Mix-based Data Augmentation Methods in Physiological Time Series Data,

P. Guo, H. Yang, and A. Sano, "Empirical Study of Mix-based Data Augmentation Methods in Physiological Time Series Data," 2023 IEEE 11th International Conference on Healthcare Informatics (ICHI), pp. 206-213, 2023

2023
[12]

Comparative Analysis of Data Augmentation for Clinical ECG Classification with STAR,

N. Nemati, "Comparative Analysis of Data Augmentation for Clinical ECG Classification with STAR," in medRxiv, 2025

2025
[13]

Optimizing Electrocardiogram Signal Augmentation for Realistic Synthetic Data in Deep Learning Model,

M. F. Safdar, P. Pałka, A. A. Faresi, and R. M. Nowak, "Optimizing Electrocardiogram Signal Augmentation for Realistic Synthetic Data in Deep Learning Model," in 2024 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SP A), 25-27 Sept. 2024 2024, pp. 54-59, doi: 10.23919/SPA61993.2024.10715629

work page doi:10.23919/spa61993.2024.10715629 2024
[14]

Electroencephalographic Signal Data Augmentation Based on Improved Generative Adversarial Network,

X. Du et al., "Electroencephalographic Signal Data Augmentation Based on Improved Generative Adversarial Network," (in eng), Brain Sci, vol. 14, no. 4, Apr 9 2024, doi: 10.3390/brainsci14040367

work page doi:10.3390/brainsci14040367 2024
[15]

Neurophysiological data augmentation for EEG-fNIRS multimodal features based on a denoising diffusion probabilistic model,

L. Chen et al., "Neurophysiological data augmentation for EEG-fNIRS multimodal features based on a denoising diffusion probabilistic model," (in eng), Comput Methods Programs Biomed, vol. 261, p. 108594, Apr 2025, doi: 10.1016/j.cmpb.2025.108594

work page doi:10.1016/j.cmpb.2025.108594 2025
[16]

Towards physiology-informed data augmentation for EEG-based BCIs,

O. Zlatov and B. Blankertz, "Towards physiology-informed data augmentation for EEG-based BCIs," ArXiv, vol. abs/2203.14392, 2022

arXiv 2022
[17]

sEMG Signal Generation for Data Augmentation Using Time Series Transformer Based Conditional GAN,

C. Nasrallah, S. Boudaoud, J. Laforet, E. Chazard, J. B. Beuscart, and D. Istrate, "sEMG Signal Generation for Data Augmentation Using Time Series Transformer Based Conditional GAN," in 2023 Seventh International Conference on Advances in Biomedical Engineering (ICABME), 12-13 Oct. 2023 2023, pp. 137-141, doi: 10.1109/ICABME59496.2023.10293077

work page doi:10.1109/icabme59496.2023.10293077 2023
[18]

COA-HAR: Exploring contrastive online test-time adaptation for wearable sensor-based human activity recognition using sensor data augmentation,

V . Fortes Rey, P. M. Bressane Rezende, B. Zhou, S. Suh, and P. Lukowicz, "COA-HAR: Exploring contrastive online test-time adaptation for wearable sensor-based human activity recognition using sensor data augmentation," Expert Systems with Applications, vol. 297, p. 129288, 2026/02/01/ 2026, doi: https://doi.org/10.1016/j.eswa.2025.129288

work page doi:10.1016/j.eswa.2025.129288 2026
[19]

BayTTA: Uncertainty-aware medical image classification with optimized test-time augmentation using Bayesian model averaging,

Z. Sherkatghanad, M. Abdar, M. Bakhtyari, P. Pławiak, and V . Makarenkov, "BayTTA: Uncertainty-aware medical image classification with optimized test-time augmentation using Bayesian model averaging," Knowledge-Based Systems, vol. 327, p. 114123, 2025/10/09/ 2025, doi: https://doi.org/10.1016/j.knosys.2025.114123

work page doi:10.1016/j.knosys.2025.114123 2025
[20]

Efficient improvement of classification accuracy via selective test-time augmentation,

J. Son and S. Kang, "Efficient improvement of classification accuracy via selective test-time augmentation," Information Sciences, vol. 642, p. 119148, 2023/09/01/ 2023, doi: https://doi.org/10.1016/j.ins.2023.119148

work page doi:10.1016/j.ins.2023.119148 2023
[21]

NeXt-TDNN: Modernizing Multi-Scale Temporal Convolution Backbone for Speaker Verification , booktitle =

S. Pan et al., "Efficient Learning on Successive Test Time Augmentation," in ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 14-19 April 2024 2024, pp. 3365-3369, doi: 10.1109/ICASSP48485.2024.10448390

work page doi:10.1109/icassp48485.2024.10448390 2024
[22]

Multi-task Self- Supervised Learning for Human Activity Detection,

A. Saeed, T. Ozcelebi, and J. Lukkien, "Multi-task Self- Supervised Learning for Human Activity Detection," Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., vol. 3, no. 2, p. Article 61, 2019, doi: 10.1145/3328932

work page doi:10.1145/3328932 2019
[23]

Sparse learned kernels for interpretable and efficient medical time series processing,

S. F. Chen, Z. Guo, C. Ding, X. Hu, and C. Rudin, "Sparse learned kernels for interpretable and efficient medical time series processing," (in eng), Nat Mach Intell, vol. 6, no. 10, pp. 1132-1144, Oct 2024, doi: 10.1038/s42256-024-00898-4

work page doi:10.1038/s42256-024-00898-4 2024
[24]

Deep learning approaches for plethysmography signal quality assessment in the presence of atrial fibrillation,

T. Pereira et al., "Deep learning approaches for plethysmography signal quality assessment in the presence of atrial fibrillation," (in eng), Physiol Meas, vol. 40, no. 12, p. 125002, Dec 27 2019, doi: 10.1088/1361-6579/ab5b84

work page doi:10.1088/1361-6579/ab5b84 2019
[25]

Log-Spectral Matching GAN: PPG-Based Atrial Fibrillation Detection can be Enhanced by GAN-Based Data Augmentation With Integration of Spectral Loss,

C. Ding et al., "Log-Spectral Matching GAN: PPG-Based Atrial Fibrillation Detection can be Enhanced by GAN-Based Data Augmentation With Integration of Spectral Loss," (in eng), IEEE J Biomed Health Inform, vol. 27, no. 3, pp. 1331- 1341, Mar 2023, doi: 10.1109/jbhi.2023.3234557

work page doi:10.1109/jbhi.2023.3234557 2023
[26]

Improving atrial fibrillation detection using a shared latent space for ECG and PPG signals,

Z. Guo et al., "Improving atrial fibrillation detection using a shared latent space for ECG and PPG signals," Harvard Data Science Review, vol. 7, no. 1, 2025/1/30 2025, doi: 10.1162/99608f92.9e63a630

work page doi:10.1162/99608f92.9e63a630 2025
[27]

C. Ding et al., "Learning From Alarms: A Robust Learning Approach for Accurate Photoplethysmography-Based Atrial Fibrillation Detection Using Eight Million Samples Labeled With Imprecise Arrhythmia Alarms," (in eng), IEEE J Biomed Health Inform, vol. 28, no. 5, pp. 2650-2661, May 2024, doi: 10.1109/jbhi.2024.3360952

work page doi:10.1109/jbhi.2024.3360952 2024
[28]

Multi-task deep learning for cardiac rhythm detection in wearable devices,

J. Torres-Soto and E. A. Ashley, "Multi-task deep learning for cardiac rhythm detection in wearable devices," npj Digital Medicine, vol. 3, no. 1, p. 116, 2020/09/09 2020, doi: 10.1038/s41746-020-00320-4

work page doi:10.1038/s41746-020-00320-4 2020
[29]

A deep learning approach to monitoring and detecting atrial fibrillation using wearable technology,

S. P. Shashikumar, A. J. Shah, Q. Li, G. D. Clifford, and S. Nemati, "A deep learning approach to monitoring and detecting atrial fibrillation using wearable technology," in 2017 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), 16-19 Feb. 2017 2017, pp. 141-144, doi: 10.1109/BHI.2017.7897225

work page doi:10.1109/bhi.2017.7897225 2017
[30]

et al.Effect of dopants and surface ox- ides on the electronic properties of dislocations in p-type SiC

Z. Chen et al., "GPT-PPG: a GPT-based foundation model for photoplethysmography signals," Physiological Measurement, vol. 46, no. 5, p. 055004, 2025/06/10 2025, doi: 10.1088/1361- 6579/add988

work page doi:10.1088/1361- 2025
[31]

SQUWA: Signal Quality Aware DNN Architecture for Enhanced Accuracy in Atrial Fibrillation Detection from Noisy PPG Signals,

R. Yan et al., "SQUWA: Signal Quality Aware DNN Architecture for Enhanced Accuracy in Atrial Fibrillation Detection from Noisy PPG Signals," ed: arXiv, 2024

2024
[32]

org/10.1109/CVPR.2016.90

K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 27-30 June 2016 2016, pp. 770-778, doi: 10.1109/CVPR.2016.90

work page doi:10.1109/cvpr.2016.90 2016
[33]

Understanding how dimension reduction tools work: an empirical approach to deciphering t-SNE, UMAP, TriMap, and PaCMAP for data visualization,

Y . Wang, H. Huang, C. Rudin, and Y . Shaposhnik, "Understanding how dimension reduction tools work: an empirical approach to deciphering t-SNE, UMAP, TriMap, and PaCMAP for data visualization," J. Mach. Learn. Res., vol. 22, no. 1, p. Article 201, 2021

2021
[34]

Atrial Fibrillation Detection from Wrist Photoplethysmography Signals Using Smartwatches,

S. K. Bashar et al., "Atrial Fibrillation Detection from Wrist Photoplethysmography Signals Using Smartwatches," Scientific Reports, vol. 9, no. 1, p. 15054, 2019/10/21 2019, doi: 10.1038/s41598-019-49092-2

work page doi:10.1038/s41598-019-49092-2 2019
[35]

J. Bacevicius et al., "High Specificity Wearable Device With Photoplethysmography and Six-Lead Electrocardiography for Atrial Fibrillation Detection Challenged by Frequent Premature Contractions: DoubleCheck-AF," (in eng), Front Cardiovasc Med, vol. 9, p. 869730, 2022, doi: 10.3389/fcvm.2022.869730

work page doi:10.3389/fcvm.2022.869730 2022
[36]

Using Apple Watch for Arrhythmia Detection,

I. Apple, "Using Apple Watch for Arrhythmia Detection," Apple Inc., Cupertino, CA, 2020-12 2020. [Online]. Available: https://www.apple.com/healthcare/docs/site/Apple_Watch_Arr hythmia_Detection.pdf

2020
[37]

WildPPG: A Real- World PPG Dataset of Long Continuous Recordings,

B. U. Demirel, C. Holz, and M. Meier, "WildPPG: A Real- World PPG Dataset of Long Continuous Recordings," presented at the Advances in Neural Information Processing Systems 37, 2024. [Online]. Available: http://dx.doi.org/10.52202/079017-0073. Journal XX (XXXX) XXXXXX Author et al 18

work page doi:10.52202/079017-0073 2024
[38]

A PPG Signal Dataset Collected in Semi-Naturalistic Settings Using Galaxy Watch,

S. Park, D. Zheng, and U. Lee, "A PPG Signal Dataset Collected in Semi-Naturalistic Settings Using Galaxy Watch," Scientific Data, vol. 12, no. 1, 2025/05/28 2025, doi: 10.1038/s41597-025-05152-z

work page doi:10.1038/s41597-025-05152-z 2025
[39]

Development of a Screening Tool for Sleep Disordered Breathing in Children Using the Phone Oximeter™,

A. Garde, P. Dehkordi, W. Karlen, D. Wensley, J. M. Ansermino, and G. A. Dumont, "Development of a Screening Tool for Sleep Disordered Breathing in Children Using the Phone Oximeter™," PLoS ONE, vol. 9, no. 11, p. e112959, 2014/11/17 2014, doi: 10.1371/journal.pone.0112959

work page doi:10.1371/journal.pone.0112959 2014
[40]

ECSMP: A dataset on emotion, cognition, sleep, and multi-model physiological signals,

Z. Gao, X. Cui, W. Wan, W. Zheng, and Z. Gu, "ECSMP: A dataset on emotion, cognition, sleep, and multi-model physiological signals," Data in Brief, vol. 39, p. 107660, 2021/12 2021, doi: 10.1016/j.dib.2021.107660

work page doi:10.1016/j.dib.2021.107660 2021
[41]

Motion Artifact Cancellation in Wearable Photoplethysmography Using Gyroscope,

H. Lee, H. Chung, and J. Lee, "Motion Artifact Cancellation in Wearable Photoplethysmography Using Gyroscope," IEEE Sensors Journal, vol. 19, no. 3, pp. 1166-1175, 2019/02/01 2019, doi: 10.1109/jsen.2018.2879970

work page doi:10.1109/jsen.2018.2879970 2019
[42]

Pulse Transit Time PPG Dataset,

P. Mehrgardt, M. Khushi, S. Poon, and A. Withana, "Pulse Transit Time PPG Dataset," PhysioNet, 2022/3 2022, doi: 10.13026/g3me-rt62

work page doi:10.13026/g3me-rt62 2022
[43]

doi: 10.3390/s19143079

A. Reiss, I. Indlekofer, P. Schmidt, and K. Van Laerhoven, "Deep PPG: Large-Scale Heart Rate Estimation with Convolutional Neural Networks," Sensors, vol. 19, no. 14, p. 3079doi: 10.3390/s19143079

work page doi:10.3390/s19143079
[44]

Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection,

P. Schmidt, A. Reiss, R. Duerichen, C. Marberger, and K. V . Laerhoven, "Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection," presented at the Proceedings of the 20th ACM International Conference on Multimodal Interaction, Boulder, CO, USA, 2018. [Online]. Available: https://doi.org/10.1145/3242969.3242985

work page doi:10.1145/3242969.3242985 2018
[45]

Compression of PPG Signals with SONG Adaptive Delta Modulation,

N. Kostov and H. Zhivomirov, "Compression of PPG Signals with SONG Adaptive Delta Modulation," in 2019 International Conference on Biomedical Innovations and Applications (BIA), 8-9 Nov. 2019 2019, pp. 1-4, doi: 10.1109/BIA48344.2019.8967471

work page doi:10.1109/bia48344.2019.8967471 2019
[46]

Simultaneous physiological measurements with five devices at different cognitive and physical loads,

M. V ollmer, D. Bläsing, J. E. Reiser, M. Nisser, and A. Buder, "Simultaneous physiological measurements with five devices at different cognitive and physical loads," PhysioNet, 2020/06 2020, doi: 10.13026/chd5-t946

work page doi:10.13026/chd5-t946 2020
[47]

AffectiveROAD system and database to assess driver's attention,

N. E. Haouij, J.-M. Poggi, S. Sevestre-Ghalila, R. Ghozi, and M. Jaïdane, "AffectiveROAD system and database to assess driver's attention," presented at the Proceedings of the 33rd Annual ACM Symposium on Applied Computing, 2018/04/09,

2018
[48]

Available: http://dx.doi.org/10.1145/3167132.3167395

[Online]. Available: http://dx.doi.org/10.1145/3167132.3167395

work page doi:10.1145/3167132.3167395
[49]

Description of a Database Containing Wrist PPG Signals Recorded during Physical Exercise with Both Accelerometer and Gyroscope Measures of Motion,

D. Jarchi and A. Casson, "Description of a Database Containing Wrist PPG Signals Recorded during Physical Exercise with Both Accelerometer and Gyroscope Measures of Motion," Data, vol. 2, no. 1, p. 1, 2016/12/24 2016, doi: 10.3390/data2010001

work page doi:10.3390/data2010001 2016
[51]

Photoplythesmogram (PPG) Signal Reliability Analysis in a Wearable Sensor-Kit,

D. Alabed, "Photoplythesmogram (PPG) Signal Reliability Analysis in a Wearable Sensor-Kit," Purdue University, 2023. [Online]. Available: https://doi.org/10.25394/PGS.8040773

work page doi:10.25394/pgs.8040773 2023
[52]

Acquiring Wearable Photoplethysmography Data in Daily Life: The PPG Diary Pilot Study,

P. H. Charlton, P. Kyriacou, J. Mant, and J. Alastruey, "Acquiring Wearable Photoplethysmography Data in Daily Life: The PPG Diary Pilot Study," presented at the 7th International Electronic Conference on Sensors and Applications, 2020/11/14, 2020. [Online]. Available: http://dx.doi.org/10.3390/ecsa-7-08233

work page doi:10.3390/ecsa-7-08233 2020
[53]

Picard, Elias Vyzas, and Jennifer Healey

R. W. Picard, E. Vyzas, and J. Healey, "Toward machine emotional intelligence: analysis of affective physiological state," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 23, no. 10, pp. 1175-1191, 2001, doi: 10.1109/34.954607

work page doi:10.1109/34.954607 2001
[54]

OpenOximetry Repository,

N. Fong, M. Lipnick, P. Bickler, J. Feiner, and T. Law, "OpenOximetry Repository," PhysioNet, 2024/02 2024, doi: 10.13026/cc78-ad74. Journal XX (XXXX) XXXXXX Author et al 19 Appendix A.1 Adding Noise Introducing controlled levels of synthetic noise during augmentation can simulate these real -world distortions, enabling the model to learn to focus on the ...

work page doi:10.13026/cc78-ad74 2024
[55]

https://zenodo.org/records/14635823 3 Development of a Screening Tool for Sleep Disordered Breathing in Children Using the Phone Oximeter

arXiv
[56]

https://figshare.com/articles/dataset/Develop ment_of_a_Screening_Tool_for_Sleep_Disor dered_Breathing_in_Children_Using_the_Ph one_Oximeter/1209662/6 4 ECSMP: A Dataset on Emotion, Cognition, Sleep, and Multi-model Physiological Signals

arXiv
[57]

https://data.mendeley.com/datasets/vn5nknh3 mn/2 5 Motion Artifact Cancellation in Wearable Photoplethysmography Using Gyroscope
[58]

https://github.com/hooseok/gyro_acc_ppg 6 Pulse Transit Time PPG Dataset [42] https://physionet.org/content/pulse-transit- time-ppg/1.0.0/ 7 PPG-DaLiA [43] https://archive.ics.uci.edu/dataset/495/ppg+d alia 8 WESAD (Wearable Stress and Affect Detection) [44] https://archive.ics.uci.edu/dataset/465/wesad +wearable+stress+and+affect+detection 9 iAMwell: ECG...

arXiv
[59]

https://physionet.org/content/simultaneous- measurements/1.0.0/ 11 AffectiveROAD system and database to assess driver's attention [47] https://www.media.mit.edu/groups/affective- computing/data/ 12 WRIST: Wrist PPG During Exercise [48] https://physionet.org/content/wrist/1.0.0/ 13 Dataset from PPG wireless sensor for activity monitoring [49] https://doi.o...

work page doi:10.1016/j.dib.2019.105044 2019
[60]

(A.5) be a linear classifier with weight vector 𝑤 ∈ ℝ𝑑 and bias 𝑏 ∈ ℝ

https://purr.purdue.edu/publications/3180/1 15 PPG Diary 1 Study Data [51] https://zenodo.org/records/5211472 16 Eight-Emotion Sentics Data [52] https://www.media.mit.edu/groups/affective- computing/data/ 17 OpenOximetry Repository [53] https://physionet.org/content/openox- repo/1.0.0/ https://openoximetry.org/data-repository/ A.6 Theoretical Basis of Dec...

arXiv

[1] [1]

Understanding Test-Time Augmentation,

M. Kimura, "Understanding Test-Time Augmentation," in Neural Information Processing, Cham, T. Mantoro, M. Lee, M. A. Ayu, K. W. Wong, and A. N. Hidayanto, Eds., 2021// 2021: Springer International Publishing, pp. 558-569

2021

[2] [2]

A survey on Image Data Augmentation for Deep Learning,

C. Shorten and T. M. Khoshgoftaar, "A survey on Image Data Augmentation for Deep Learning," Journal of Big Data, vol. 6, no. 1, p. 60, 2019/07/06 2019, doi: 10.1186/s40537-019-0197- 0

work page doi:10.1186/s40537-019-0197- 2019

[3] [3]

Data augmentation of wearable sensor data for parkinson’s disease monitoring using convolutional neural networks,

T. T. Um et al., "Data augmentation of wearable sensor data for parkinson’s disease monitoring using convolutional neural networks," presented at the Proceedings of the 19th ACM International Conference on Multimodal Interaction, Glasgow, UK, 2017. [Online]. Available: https://doi.org/10.1145/3136755.3136817

work page doi:10.1145/3136755.3136817 2017

[4] [4]

A Multimodal-Driven Fusion Data Augmentation Framework for Emotion Recognition,

A. Li, M. Wu, R. Ouyang, Y . Wang, F. Li, and Z. Lv, "A Multimodal-Driven Fusion Data Augmentation Framework for Emotion Recognition," IEEE Transactions on Artificial Intelligence, vol. 6, no. 8, pp. 2083-2097, 2025, doi: 10.1109/TAI.2025.3537965

work page doi:10.1109/tai.2025.3537965 2083

[5] [5]

A novel data augmentation strategy for robust deep learning classification of biomedical time-series data: Application to ECG and EEG analysis,

M. Guhdar, R. J. Mstafa, and A. O. Mohammed, "A novel data augmentation strategy for robust deep learning classification of biomedical time-series data: Application to ECG and EEG analysis," arXiv [eess.SP], 2025/7/16 2025. [Online]. Available: http://arxiv.org/abs/2507.12645

Pith/arXiv arXiv 2025

[6] [6]

In: Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, volume 2022-July

J. An, R. E. Gregg, and S. Borhani, "Effective Data Augmentation, Filters, and Automation Techniques for Automatic 12-Lead ECG Classification Using Deep Residual Neural Networks," in 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 11-15 July 2022 2022, pp. 1283-1287, doi: 10.1109/EMBC48229.2022.9871654

work page doi:10.1109/embc48229.2022.9871654 2022

[7] [7]

Explanations of Augmentation Methods for Deep Learning ECG Classification,

N. S. P. Balasubramanian and S. Dakshit, "Explanations of Augmentation Methods for Deep Learning ECG Classification," in Artificial Intelligence in Medicine, Cham, J. Finkelstein, R. Moskovitch, and E. Parimbelli, Eds., 2024// 2024: Springer Nature Switzerland, pp. 277-287

2024

[8] [8]

A novel data augmentation method to enhance deep neural networks for detection of atrial fibrillation,

P. Cao et al., "A novel data augmentation method to enhance deep neural networks for detection of atrial fibrillation," Biomedical Signal Processing and Control, vol. 56, p. 101675, 2020/02/01/ 2020, doi: https://doi.org/10.1016/j.bspc.2019.101675

work page doi:10.1016/j.bspc.2019.101675 2020

[9] [9]

Emotion Detection Using Physiological Signals,

A. Dessai and H. Virani, "Emotion Detection Using Physiological Signals," in 2021 International Conference on Electrical, Computer and Energy Technologies (ICECET), 9- 10 Dec. 2021 2021, pp. 1-4, doi: 10.1109/ICECET52533.2021.9698729. Journal XX (XXXX) XXXXXX Author et al 17

work page doi:10.1109/icecet52533.2021.9698729 2021

[10] [10]

Data Augmentation for 12-Lead ECG Beat Classification,

E. Do, J. Boynton, B. S. Lee, and D. Lustgarten, "Data Augmentation for 12-Lead ECG Beat Classification," SN Computer Science, vol. 3, no. 1, p. 70, 2021/11/19 2021, doi: 10.1007/s42979-021-00924-x

work page doi:10.1007/s42979-021-00924-x 2021

[11] [11]

Empirical Study of Mix-based Data Augmentation Methods in Physiological Time Series Data,

P. Guo, H. Yang, and A. Sano, "Empirical Study of Mix-based Data Augmentation Methods in Physiological Time Series Data," 2023 IEEE 11th International Conference on Healthcare Informatics (ICHI), pp. 206-213, 2023

2023

[12] [12]

Comparative Analysis of Data Augmentation for Clinical ECG Classification with STAR,

N. Nemati, "Comparative Analysis of Data Augmentation for Clinical ECG Classification with STAR," in medRxiv, 2025

2025

[13] [13]

Optimizing Electrocardiogram Signal Augmentation for Realistic Synthetic Data in Deep Learning Model,

M. F. Safdar, P. Pałka, A. A. Faresi, and R. M. Nowak, "Optimizing Electrocardiogram Signal Augmentation for Realistic Synthetic Data in Deep Learning Model," in 2024 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SP A), 25-27 Sept. 2024 2024, pp. 54-59, doi: 10.23919/SPA61993.2024.10715629

work page doi:10.23919/spa61993.2024.10715629 2024

[14] [14]

Electroencephalographic Signal Data Augmentation Based on Improved Generative Adversarial Network,

X. Du et al., "Electroencephalographic Signal Data Augmentation Based on Improved Generative Adversarial Network," (in eng), Brain Sci, vol. 14, no. 4, Apr 9 2024, doi: 10.3390/brainsci14040367

work page doi:10.3390/brainsci14040367 2024

[15] [15]

Neurophysiological data augmentation for EEG-fNIRS multimodal features based on a denoising diffusion probabilistic model,

L. Chen et al., "Neurophysiological data augmentation for EEG-fNIRS multimodal features based on a denoising diffusion probabilistic model," (in eng), Comput Methods Programs Biomed, vol. 261, p. 108594, Apr 2025, doi: 10.1016/j.cmpb.2025.108594

work page doi:10.1016/j.cmpb.2025.108594 2025

[16] [16]

Towards physiology-informed data augmentation for EEG-based BCIs,

O. Zlatov and B. Blankertz, "Towards physiology-informed data augmentation for EEG-based BCIs," ArXiv, vol. abs/2203.14392, 2022

arXiv 2022

[17] [17]

sEMG Signal Generation for Data Augmentation Using Time Series Transformer Based Conditional GAN,

C. Nasrallah, S. Boudaoud, J. Laforet, E. Chazard, J. B. Beuscart, and D. Istrate, "sEMG Signal Generation for Data Augmentation Using Time Series Transformer Based Conditional GAN," in 2023 Seventh International Conference on Advances in Biomedical Engineering (ICABME), 12-13 Oct. 2023 2023, pp. 137-141, doi: 10.1109/ICABME59496.2023.10293077

work page doi:10.1109/icabme59496.2023.10293077 2023

[18] [18]

COA-HAR: Exploring contrastive online test-time adaptation for wearable sensor-based human activity recognition using sensor data augmentation,

V . Fortes Rey, P. M. Bressane Rezende, B. Zhou, S. Suh, and P. Lukowicz, "COA-HAR: Exploring contrastive online test-time adaptation for wearable sensor-based human activity recognition using sensor data augmentation," Expert Systems with Applications, vol. 297, p. 129288, 2026/02/01/ 2026, doi: https://doi.org/10.1016/j.eswa.2025.129288

work page doi:10.1016/j.eswa.2025.129288 2026

[19] [19]

BayTTA: Uncertainty-aware medical image classification with optimized test-time augmentation using Bayesian model averaging,

Z. Sherkatghanad, M. Abdar, M. Bakhtyari, P. Pławiak, and V . Makarenkov, "BayTTA: Uncertainty-aware medical image classification with optimized test-time augmentation using Bayesian model averaging," Knowledge-Based Systems, vol. 327, p. 114123, 2025/10/09/ 2025, doi: https://doi.org/10.1016/j.knosys.2025.114123

work page doi:10.1016/j.knosys.2025.114123 2025

[20] [20]

Efficient improvement of classification accuracy via selective test-time augmentation,

J. Son and S. Kang, "Efficient improvement of classification accuracy via selective test-time augmentation," Information Sciences, vol. 642, p. 119148, 2023/09/01/ 2023, doi: https://doi.org/10.1016/j.ins.2023.119148

work page doi:10.1016/j.ins.2023.119148 2023

[21] [21]

NeXt-TDNN: Modernizing Multi-Scale Temporal Convolution Backbone for Speaker Verification , booktitle =

S. Pan et al., "Efficient Learning on Successive Test Time Augmentation," in ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 14-19 April 2024 2024, pp. 3365-3369, doi: 10.1109/ICASSP48485.2024.10448390

work page doi:10.1109/icassp48485.2024.10448390 2024

[22] [22]

Multi-task Self- Supervised Learning for Human Activity Detection,

A. Saeed, T. Ozcelebi, and J. Lukkien, "Multi-task Self- Supervised Learning for Human Activity Detection," Proc. ACM Interact. Mob. Wearable Ubiquitous Technol., vol. 3, no. 2, p. Article 61, 2019, doi: 10.1145/3328932

work page doi:10.1145/3328932 2019

[23] [23]

Sparse learned kernels for interpretable and efficient medical time series processing,

S. F. Chen, Z. Guo, C. Ding, X. Hu, and C. Rudin, "Sparse learned kernels for interpretable and efficient medical time series processing," (in eng), Nat Mach Intell, vol. 6, no. 10, pp. 1132-1144, Oct 2024, doi: 10.1038/s42256-024-00898-4

work page doi:10.1038/s42256-024-00898-4 2024

[24] [24]

Deep learning approaches for plethysmography signal quality assessment in the presence of atrial fibrillation,

T. Pereira et al., "Deep learning approaches for plethysmography signal quality assessment in the presence of atrial fibrillation," (in eng), Physiol Meas, vol. 40, no. 12, p. 125002, Dec 27 2019, doi: 10.1088/1361-6579/ab5b84

work page doi:10.1088/1361-6579/ab5b84 2019

[25] [25]

Log-Spectral Matching GAN: PPG-Based Atrial Fibrillation Detection can be Enhanced by GAN-Based Data Augmentation With Integration of Spectral Loss,

C. Ding et al., "Log-Spectral Matching GAN: PPG-Based Atrial Fibrillation Detection can be Enhanced by GAN-Based Data Augmentation With Integration of Spectral Loss," (in eng), IEEE J Biomed Health Inform, vol. 27, no. 3, pp. 1331- 1341, Mar 2023, doi: 10.1109/jbhi.2023.3234557

work page doi:10.1109/jbhi.2023.3234557 2023

[26] [26]

Improving atrial fibrillation detection using a shared latent space for ECG and PPG signals,

Z. Guo et al., "Improving atrial fibrillation detection using a shared latent space for ECG and PPG signals," Harvard Data Science Review, vol. 7, no. 1, 2025/1/30 2025, doi: 10.1162/99608f92.9e63a630

work page doi:10.1162/99608f92.9e63a630 2025

[27] [27]

C. Ding et al., "Learning From Alarms: A Robust Learning Approach for Accurate Photoplethysmography-Based Atrial Fibrillation Detection Using Eight Million Samples Labeled With Imprecise Arrhythmia Alarms," (in eng), IEEE J Biomed Health Inform, vol. 28, no. 5, pp. 2650-2661, May 2024, doi: 10.1109/jbhi.2024.3360952

work page doi:10.1109/jbhi.2024.3360952 2024

[28] [28]

Multi-task deep learning for cardiac rhythm detection in wearable devices,

J. Torres-Soto and E. A. Ashley, "Multi-task deep learning for cardiac rhythm detection in wearable devices," npj Digital Medicine, vol. 3, no. 1, p. 116, 2020/09/09 2020, doi: 10.1038/s41746-020-00320-4

work page doi:10.1038/s41746-020-00320-4 2020

[29] [29]

A deep learning approach to monitoring and detecting atrial fibrillation using wearable technology,

S. P. Shashikumar, A. J. Shah, Q. Li, G. D. Clifford, and S. Nemati, "A deep learning approach to monitoring and detecting atrial fibrillation using wearable technology," in 2017 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), 16-19 Feb. 2017 2017, pp. 141-144, doi: 10.1109/BHI.2017.7897225

work page doi:10.1109/bhi.2017.7897225 2017

[30] [30]

et al.Effect of dopants and surface ox- ides on the electronic properties of dislocations in p-type SiC

Z. Chen et al., "GPT-PPG: a GPT-based foundation model for photoplethysmography signals," Physiological Measurement, vol. 46, no. 5, p. 055004, 2025/06/10 2025, doi: 10.1088/1361- 6579/add988

work page doi:10.1088/1361- 2025

[31] [31]

SQUWA: Signal Quality Aware DNN Architecture for Enhanced Accuracy in Atrial Fibrillation Detection from Noisy PPG Signals,

R. Yan et al., "SQUWA: Signal Quality Aware DNN Architecture for Enhanced Accuracy in Atrial Fibrillation Detection from Noisy PPG Signals," ed: arXiv, 2024

2024

[32] [32]

org/10.1109/CVPR.2016.90

K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 27-30 June 2016 2016, pp. 770-778, doi: 10.1109/CVPR.2016.90

work page doi:10.1109/cvpr.2016.90 2016

[33] [33]

Understanding how dimension reduction tools work: an empirical approach to deciphering t-SNE, UMAP, TriMap, and PaCMAP for data visualization,

Y . Wang, H. Huang, C. Rudin, and Y . Shaposhnik, "Understanding how dimension reduction tools work: an empirical approach to deciphering t-SNE, UMAP, TriMap, and PaCMAP for data visualization," J. Mach. Learn. Res., vol. 22, no. 1, p. Article 201, 2021

2021

[34] [34]

Atrial Fibrillation Detection from Wrist Photoplethysmography Signals Using Smartwatches,

S. K. Bashar et al., "Atrial Fibrillation Detection from Wrist Photoplethysmography Signals Using Smartwatches," Scientific Reports, vol. 9, no. 1, p. 15054, 2019/10/21 2019, doi: 10.1038/s41598-019-49092-2

work page doi:10.1038/s41598-019-49092-2 2019

[35] [35]

J. Bacevicius et al., "High Specificity Wearable Device With Photoplethysmography and Six-Lead Electrocardiography for Atrial Fibrillation Detection Challenged by Frequent Premature Contractions: DoubleCheck-AF," (in eng), Front Cardiovasc Med, vol. 9, p. 869730, 2022, doi: 10.3389/fcvm.2022.869730

work page doi:10.3389/fcvm.2022.869730 2022

[36] [36]

Using Apple Watch for Arrhythmia Detection,

I. Apple, "Using Apple Watch for Arrhythmia Detection," Apple Inc., Cupertino, CA, 2020-12 2020. [Online]. Available: https://www.apple.com/healthcare/docs/site/Apple_Watch_Arr hythmia_Detection.pdf

2020

[37] [37]

WildPPG: A Real- World PPG Dataset of Long Continuous Recordings,

B. U. Demirel, C. Holz, and M. Meier, "WildPPG: A Real- World PPG Dataset of Long Continuous Recordings," presented at the Advances in Neural Information Processing Systems 37, 2024. [Online]. Available: http://dx.doi.org/10.52202/079017-0073. Journal XX (XXXX) XXXXXX Author et al 18

work page doi:10.52202/079017-0073 2024

[38] [38]

A PPG Signal Dataset Collected in Semi-Naturalistic Settings Using Galaxy Watch,

S. Park, D. Zheng, and U. Lee, "A PPG Signal Dataset Collected in Semi-Naturalistic Settings Using Galaxy Watch," Scientific Data, vol. 12, no. 1, 2025/05/28 2025, doi: 10.1038/s41597-025-05152-z

work page doi:10.1038/s41597-025-05152-z 2025

[39] [39]

Development of a Screening Tool for Sleep Disordered Breathing in Children Using the Phone Oximeter™,

A. Garde, P. Dehkordi, W. Karlen, D. Wensley, J. M. Ansermino, and G. A. Dumont, "Development of a Screening Tool for Sleep Disordered Breathing in Children Using the Phone Oximeter™," PLoS ONE, vol. 9, no. 11, p. e112959, 2014/11/17 2014, doi: 10.1371/journal.pone.0112959

work page doi:10.1371/journal.pone.0112959 2014

[40] [40]

ECSMP: A dataset on emotion, cognition, sleep, and multi-model physiological signals,

Z. Gao, X. Cui, W. Wan, W. Zheng, and Z. Gu, "ECSMP: A dataset on emotion, cognition, sleep, and multi-model physiological signals," Data in Brief, vol. 39, p. 107660, 2021/12 2021, doi: 10.1016/j.dib.2021.107660

work page doi:10.1016/j.dib.2021.107660 2021

[41] [41]

Motion Artifact Cancellation in Wearable Photoplethysmography Using Gyroscope,

H. Lee, H. Chung, and J. Lee, "Motion Artifact Cancellation in Wearable Photoplethysmography Using Gyroscope," IEEE Sensors Journal, vol. 19, no. 3, pp. 1166-1175, 2019/02/01 2019, doi: 10.1109/jsen.2018.2879970

work page doi:10.1109/jsen.2018.2879970 2019

[42] [42]

Pulse Transit Time PPG Dataset,

P. Mehrgardt, M. Khushi, S. Poon, and A. Withana, "Pulse Transit Time PPG Dataset," PhysioNet, 2022/3 2022, doi: 10.13026/g3me-rt62

work page doi:10.13026/g3me-rt62 2022

[43] [43]

doi: 10.3390/s19143079

A. Reiss, I. Indlekofer, P. Schmidt, and K. Van Laerhoven, "Deep PPG: Large-Scale Heart Rate Estimation with Convolutional Neural Networks," Sensors, vol. 19, no. 14, p. 3079doi: 10.3390/s19143079

work page doi:10.3390/s19143079

[44] [44]

Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection,

P. Schmidt, A. Reiss, R. Duerichen, C. Marberger, and K. V . Laerhoven, "Introducing WESAD, a Multimodal Dataset for Wearable Stress and Affect Detection," presented at the Proceedings of the 20th ACM International Conference on Multimodal Interaction, Boulder, CO, USA, 2018. [Online]. Available: https://doi.org/10.1145/3242969.3242985

work page doi:10.1145/3242969.3242985 2018

[45] [45]

Compression of PPG Signals with SONG Adaptive Delta Modulation,

N. Kostov and H. Zhivomirov, "Compression of PPG Signals with SONG Adaptive Delta Modulation," in 2019 International Conference on Biomedical Innovations and Applications (BIA), 8-9 Nov. 2019 2019, pp. 1-4, doi: 10.1109/BIA48344.2019.8967471

work page doi:10.1109/bia48344.2019.8967471 2019

[46] [46]

Simultaneous physiological measurements with five devices at different cognitive and physical loads,

M. V ollmer, D. Bläsing, J. E. Reiser, M. Nisser, and A. Buder, "Simultaneous physiological measurements with five devices at different cognitive and physical loads," PhysioNet, 2020/06 2020, doi: 10.13026/chd5-t946

work page doi:10.13026/chd5-t946 2020

[47] [47]

AffectiveROAD system and database to assess driver's attention,

N. E. Haouij, J.-M. Poggi, S. Sevestre-Ghalila, R. Ghozi, and M. Jaïdane, "AffectiveROAD system and database to assess driver's attention," presented at the Proceedings of the 33rd Annual ACM Symposium on Applied Computing, 2018/04/09,

2018

[48] [48]

Available: http://dx.doi.org/10.1145/3167132.3167395

[Online]. Available: http://dx.doi.org/10.1145/3167132.3167395

work page doi:10.1145/3167132.3167395

[49] [49]

Description of a Database Containing Wrist PPG Signals Recorded during Physical Exercise with Both Accelerometer and Gyroscope Measures of Motion,

D. Jarchi and A. Casson, "Description of a Database Containing Wrist PPG Signals Recorded during Physical Exercise with Both Accelerometer and Gyroscope Measures of Motion," Data, vol. 2, no. 1, p. 1, 2016/12/24 2016, doi: 10.3390/data2010001

work page doi:10.3390/data2010001 2016

[50] [51]

Photoplythesmogram (PPG) Signal Reliability Analysis in a Wearable Sensor-Kit,

D. Alabed, "Photoplythesmogram (PPG) Signal Reliability Analysis in a Wearable Sensor-Kit," Purdue University, 2023. [Online]. Available: https://doi.org/10.25394/PGS.8040773

work page doi:10.25394/pgs.8040773 2023

[51] [52]

Acquiring Wearable Photoplethysmography Data in Daily Life: The PPG Diary Pilot Study,

P. H. Charlton, P. Kyriacou, J. Mant, and J. Alastruey, "Acquiring Wearable Photoplethysmography Data in Daily Life: The PPG Diary Pilot Study," presented at the 7th International Electronic Conference on Sensors and Applications, 2020/11/14, 2020. [Online]. Available: http://dx.doi.org/10.3390/ecsa-7-08233

work page doi:10.3390/ecsa-7-08233 2020

[52] [53]

Picard, Elias Vyzas, and Jennifer Healey

R. W. Picard, E. Vyzas, and J. Healey, "Toward machine emotional intelligence: analysis of affective physiological state," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 23, no. 10, pp. 1175-1191, 2001, doi: 10.1109/34.954607

work page doi:10.1109/34.954607 2001

[53] [54]

OpenOximetry Repository,

N. Fong, M. Lipnick, P. Bickler, J. Feiner, and T. Law, "OpenOximetry Repository," PhysioNet, 2024/02 2024, doi: 10.13026/cc78-ad74. Journal XX (XXXX) XXXXXX Author et al 19 Appendix A.1 Adding Noise Introducing controlled levels of synthetic noise during augmentation can simulate these real -world distortions, enabling the model to learn to focus on the ...

work page doi:10.13026/cc78-ad74 2024

[54] [55]

https://zenodo.org/records/14635823 3 Development of a Screening Tool for Sleep Disordered Breathing in Children Using the Phone Oximeter

arXiv

[55] [56]

https://figshare.com/articles/dataset/Develop ment_of_a_Screening_Tool_for_Sleep_Disor dered_Breathing_in_Children_Using_the_Ph one_Oximeter/1209662/6 4 ECSMP: A Dataset on Emotion, Cognition, Sleep, and Multi-model Physiological Signals

arXiv

[56] [57]

https://data.mendeley.com/datasets/vn5nknh3 mn/2 5 Motion Artifact Cancellation in Wearable Photoplethysmography Using Gyroscope

[57] [58]

https://github.com/hooseok/gyro_acc_ppg 6 Pulse Transit Time PPG Dataset [42] https://physionet.org/content/pulse-transit- time-ppg/1.0.0/ 7 PPG-DaLiA [43] https://archive.ics.uci.edu/dataset/495/ppg+d alia 8 WESAD (Wearable Stress and Affect Detection) [44] https://archive.ics.uci.edu/dataset/465/wesad +wearable+stress+and+affect+detection 9 iAMwell: ECG...

arXiv

[58] [59]

https://physionet.org/content/simultaneous- measurements/1.0.0/ 11 AffectiveROAD system and database to assess driver's attention [47] https://www.media.mit.edu/groups/affective- computing/data/ 12 WRIST: Wrist PPG During Exercise [48] https://physionet.org/content/wrist/1.0.0/ 13 Dataset from PPG wireless sensor for activity monitoring [49] https://doi.o...

work page doi:10.1016/j.dib.2019.105044 2019

[59] [60]

(A.5) be a linear classifier with weight vector 𝑤 ∈ ℝ𝑑 and bias 𝑏 ∈ ℝ

https://purr.purdue.edu/publications/3180/1 15 PPG Diary 1 Study Data [51] https://zenodo.org/records/5211472 16 Eight-Emotion Sentics Data [52] https://www.media.mit.edu/groups/affective- computing/data/ 17 OpenOximetry Repository [53] https://physionet.org/content/openox- repo/1.0.0/ https://openoximetry.org/data-repository/ A.6 Theoretical Basis of Dec...

arXiv