Pith. sign in

REVIEW 3 major objections 4 minor 16 references

Real-Time, Single-Ear, Wearable ECG Reconstruction, R-Peak Detection, and HR/HRV Monitoring

T0 review · 3 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A single-ear, dry-electrode wearable can reconstruct an ECG, detect R-peaks, and estimate HR/HRV in real time on an ultra-low-power chip, using only 36.7 µJ per inference.

desk verdict A credible embedded-systems step toward single-ear HR/HRV monitoring, but the headline accuracy numbers are inflated by test-set threshold tuning and a median-based correction that hides real HRV; the engineering deserves a referee, the evaluation needs rework. read the letter →

arxiv 2505.01738 v1 pith:SP4X73FC submitted 2025-05-03 eess.SP cs.SYeess.SY

Carlos Santos , Sebastian Frey , Andrea Cossettini , Luca Benini , Victor Kartsch This is my paper · ORCID
classification eess.SPcs.SYeess.SY
keywords ear-ECGR-peakdetectionheartratevariabilitywearableECGon-deviceinferencedryelectrodesDeepMFlow-powerembedded
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that a fully wearable, single-ear device with dry in-ear electrodes can reconstruct an ECG-like signal, detect R-peaks, and compute heart rate (HR) and heart-rate variability (HRV) in real time on an ultra-low-power processor, without any per-user calibration. That matters because it removes two obstacles to continuous at-home cardiovascular monitoring: wet electrodes and wiring between the ears. In the paper's tests, the on-device model used 36.7 µJ per inference and achieved a mean HR error of 0.49 bpm and a mean HRV error of 25.82 ms in the cross-ear configuration, with a single-ear F1-score of 0.82 for R-peak detection.

What carries the argument

The load-bearing mechanism is the DeepMF-mini network plus its postprocessing pipeline. DeepMF-mini is a 1D convolutional autoencoder (roughly 10k parameters) whose encoder is frozen after training and paired with a classifier made of transpose convolutions; it learns a latent-space representation shared between the noisy ear biopotential and the clean arm-ECG, and the classifier outputs R-peak probabilities. The rolling-window refinement evaluates each 2-second window multiple times with a 0.4 s shift, discards the least-reliable last segment, and takes element-wise maxima over the overlap, so peaks near window edges get a second chance. Finally, Algorithm 1 corrects the predicted peak sequence by adjusting to the running median of RR intervals, which is what makes the reported HRV numbers possible from noisy single-ear signals.

What would settle it

Record ear-ECG and simultaneous arm-ECG from, say, twenty subjects during rest, walking, and a protocol that induces premature ventricular contractions; compute RMSSD from the corrected peaks and compare with the arm-ECG ground truth. If the correction algorithm removes or inserts beats outside isolated missed/duplicate cases, the single-ear HRV error will climb substantially above the reported 91.57 ms as the arrhythmia or motion burden increases.

Watch

Extended reading notes

Core claim

The paper's central claim is that a compact convolutional network called DeepMF-mini, a roughly 75× smaller variant of the DeepMF architecture, can be trained once on ear biopotentials and then detect R-peaks in real time for new users without recalibration. The model is a two-stage design: an encoder-decoder first maps the ear signal to a latent representation that reconstructs the arm-measured Lead I ECG; the decoder is then replaced by a peak classifier trained on the same latent space. A rolling-window postprocessor merges overlapping predictions to catch peaks near window boundaries, and a correction algorithm that uses the running median of RR intervals removes duplicates and inserts missed beats. With all processing on the wearable, the paper reports an R-peak F1-score of 0.95 for cross-ear and 0.82 for single-ear input, HR errors of 0.49 bpm and 5.91 bpm, and HRV errors of 25.82 ms and 91.57 ms, respectively, at 36.7 µJ per inference.

Load-bearing premise

The R-peak correction step assumes that missed or duplicate detections are isolated events, so that deleting or inserting a beat according to the running median of RR intervals does not erase genuine beat-to-beat variability; if real arrhythmic or motion-induced intervals occur, the correction would bias the HRV measurement toward regularity and the reported accuracy would be optimistic.

Editorial extensions

If this is right

  • Consumer earbuds using this design could track HR and HRV continuously for about 36 hours on a 75 mAh battery, since the whole system draws about 7.6 mW.
  • Cross-ear R-peak detection (F1 = 0.95) is achieved with 75× fewer parameters than the reference DeepMF model, freeing memory and compute for audio processing on the same chip.
  • The single-ear configuration reduces hardware complexity (no wiring across the head) and still reaches HR errors around 5.91 bpm, suitable for trend monitoring rather than clinical-grade beat annotation.
  • The quantized 10 kB model runs in 1.47 ms per inference at the energy-efficient operating point, so it occupies only a small fraction of the duty cycle.
  • Because the model is trained with leave-one-subject-out, a new user would not need a personal calibration session before the device starts reporting HR/HRV.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's HRV numbers depend on Algorithm 1, which assumes missed or duplicate beats are isolated and true intervals are near whole-number multiples of the running median; if real arrhythmias or motion artifacts produce genuine short-long-short patterns, this correction would bias HRV toward regularity, so the 25.82 ms error may not hold on ambulatory or arrhythmic data.
  • The single-ear input's main weakness is the attenuated, less consistent T-wave and post-QRS signal; adding a second sensing modality already present in earbuds (e.g., PPG or an accelerometer for motion rejection) could plausibly close most of the 91.57 ms HRV gap without a major energy cost.
  • The same latent-space pipeline could be retargeted to estimate other cardiac intervals, such as QT or PR, if the ECG reconstruction quality improves; the paper already shows the reconstruction approximates the arm ECG waveform.
  • The 36.7 µJ per inference figure suggests an always-on heartbeat detector leaves ample power headroom for audio streaming on the same earbud, which is the practical path to making cardiac monitoring a background feature.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper presents DeepMF-mini, a lightweight convolutional autoencoder and R-peak classifier derived from the DeepMF architecture, integrated into BioGAP, an ultra-low-power wearable biosignal platform. The authors collect simultaneous arm-ECG and in-ear biopotential data from four healthy subjects in rest conditions and evaluate leave-one-subject-out (LOSO) R-peak detection, ECG reconstruction, and HR/HRV estimation for both single-ear and cross-ear configurations. They report a single-ear F1-score of 0.82, a cross-ear F1-score of 0.95, HR mean errors of 5.91 bpm (single-ear) and 0.49 bpm (cross-ear), HRV mean errors of 91.57 ms and 25.82 ms respectively, and an on-device inference energy of 36.7 uJ/inference with an estimated 36 h battery life. The main claims are that a subject-independent, fully wearable single-ear system can perform real-time HR/HRV monitoring at accuracy comparable to state-of-the-art cross-ear approaches.

Significance. If the reported accuracy holds under a properly unbiased evaluation, this is a useful step toward practical hearable-based cardiovascular monitoring: dry in-ear electrodes, a 10k-parameter model, on-device inference at 36.7 uJ per inference, and a concrete embedded deployment are all meaningful engineering contributions. The manuscript also ships code and data (github.com/pulp-bio/DeepMF-mini), which is a strength for reproducibility. The cross-ear HR error of 0.49 bpm and the sub-40 uJ inference energy are attractive headline numbers. However, the current evidence base is narrow: four male subjects, rest-only recordings, no confidence intervals, and an evaluation protocol that tunes the detection threshold on the test subjects. These issues directly affect the load-bearing 'subject-independent robust extraction' claim and need to be addressed before the quantitative claims can be taken at face value.

major comments (3)
  1. [Section III-G] The detection threshold is selected on the evaluation set: 'the threshold maximizing the F1-score was selected for rolling output inference across all subjects.' Because the evaluation is LOSO with only four subjects and no separate validation split is described, this is equivalent to tuning a hyperparameter on the test data. This biases the reported F1-score and all downstream HR/HRV errors upward, undermining the subject-independent claim. Please use nested LOSO (or a subject-disjoint validation split) and report metrics for the threshold chosen without access to test labels.
  2. [Section III-E, Algorithm 1] All reported HRV errors (25.82 ms cross-ear, 91.57 ms single-ear) are computed after applying Algorithm 1, which inserts or removes beats by rounding the ratio of each RR interval to the running median. This correction assumes that missed or duplicate detections are isolated and that the underlying true RR intervals are close to integer multiples of a median interval. On a four-subject, rest-only dataset with no arrhythmic or motion-corrupted intervals, the correction is not stress-tested, and it could suppress genuine beat-to-beat variability if such intervals occur. Please report both raw and corrected HR/HRV errors and validate the correction on data containing artifacts or known ectopic beats.
  3. [Section III-A and Section IV-C] The subject-independent claim rests on LOSO evaluation over only four healthy male subjects, all recorded at rest, with no confidence intervals or per-subject breakdowns. With n=4, a single subject can dominate the cross-ear HR error of 0.49 bpm, and the absence of variability measures makes it impossible to assess whether the reported accuracy is stable. Please provide per-subject results, confidence intervals, and preferably an independent test cohort or an explicit statistical statement of the precision of the reported errors.
minor comments (4)
  1. [Section I] The contributions list contains a typo: 'achieves SoA performance for HV and HVR' should presumably read 'HR and HRV'; elsewhere the paper correctly uses 'HRV'.
  2. [Table II] The reported DeepMF precision/recall values are inconsistent between the text (Section II: '95%/92%') and Table II ('91.2% / 94.9%'). Please reconcile these numbers.
  3. [Section IV-D and Table I] The relationship between the reported 36.7 uJ per inference, the 400 ms update interval, and the total system power of 7.6 mW is not derived; please state which components contribute to the 7.6 mW and how the 36 h battery estimate is obtained.
  4. [Section III-A] The dataset description reports mean age as '31.0 ± 5.1 years' with a decimal point; '31 ± 5 years' would be more natural, and the number of female participants (zero) should be stated explicitly as a limitation.

Circularity Check

1 steps flagged · score 6.0 of 10

Test-set threshold selection makes the reported subject-independent F1 and HR/HRV errors partly in-sample optima.

  1. fitted input called prediction [Section III-G (Accuracy evaluation), used for the results in Section IV-C]
    "To minimize subject bias in the LOSO evaluation, the threshold maximizing the F1-score was selected for rolling output inference across all subjects."

    The threshold is a free decision parameter chosen by maximizing F1 on the same LOSO test outputs that are later reported. No validation split is described, so 'across all subjects' covers every test fold. Consequently the reported F1 scores (0.82 single-ear, 0.95 cross-ear) and the downstream HR/HRV mean errors (0.49 bpm, 25.82 ms) are in-sample evaluations at the selected threshold, not out-of-sample subject-independent predictions. The subject-independent claim is therefore supported by a parameter fitted to the target labels, making the headline quantitative results partially forced by construction.

full rationale

The model-training derivation is otherwise self-contained: the encoder-decoder maps ear biopotentials to simultaneously recorded arm-ECG traces, the classifier is trained on binary R-peak labels derived from the arm-ECG, and the LOSO splits keep subjects disjoint. Algorithm 1 is a deterministic postprocessor using only predicted peak intervals and a running median; it is not circular because it does not use ground-truth labels, though its integer-multiple-of-median assumption can bias HRV toward regularity in arrhythmic or motion-corrupted data (a correctness risk, not circularity). The self-citations to BioGAP [14] and GAPSES [5] are platform/hardware references and are not load-bearing for the algorithmic claims. The single circular element is the test-set threshold selection in Section III-G. Because the F1-maximizing threshold is chosen on all LOSO test subjects and then the same subjects are used to report F1 and HR/HRV errors, the headline numbers are partly in-sample. This does not make the whole model equivalent to its inputs, but it materially compromises the 'subject-independent' quantitative claims.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The learned network weights are trained parameters and are not counted here. The listed preprocessing and evaluation-time choices are quantities the reported accuracy figures depend on but that are not learned from held-out data.

free parameters (3)
  • F1-maximizing detection threshold = not reported
    Selected on the LOSO evaluation set across all subjects (Section III-G), not on a held-out validation set; all precision/recall and HR/HRV metrics depend on it.
  • R-peak label generation thresholds = min peak height 0.7, min inter-peak distance 80 samples
    Used to derive ground-truth R-peak labels from arm-ECG (Section III-B); changes here change training targets and all downstream metrics.
  • Rolling window and correction hyperparameters = 2 s window, 0.4 s shift, 0.4 s end discard, 10 s HR/HRV window, minimum 4 peaks
    Hand-chosen postprocessing parameters (Sections III-E, III-F) that affect R-peak refinement and HRV computation; not optimized with a held-out set.
assumptions (4)
  • domain assumption In-ear and mastoid dry-electrode biopotentials contain ECG-related signal with sufficient SNR for R-peak detection.
    Supported by prior work [6] and [10], but the quantitative sufficiency for real-time single-ear detection is exactly what this small study tests (Section III-A).
  • domain assumption Four healthy male volunteers at rest are representative enough for a 'user-generic' subject-independent model.
    The paper trains with LOSO on four subjects and calls the approach subject-independent (Sections III-C and V); no diversity or motion conditions are included.
  • ad hoc to paper Median-RR correction (Algorithm 1) fixes only spurious missed or extra peaks and preserves true physiological RR variability.
    Algorithm 1 assumes RR intervals are integer multiples of the median; this can remove real variability and directly affects reported HRV errors (Sections III-E and IV-C).
  • standard math Standard signal processing (notch and bandpass filtering, z-score normalization) preserves R-peak timing relative to ground truth.
    Used in Section III-B; routine, but any phase distortion would shift labels for both training and evaluation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Real-Time, Single-Ear, Wearable ECG Reconstruction, R-Peak Detection, and HR/HRV Monitoring." pith.science (2026). https://pith.science/paper/SP4X73FC

@misc{pith2026250501738,
  author       = {Pith},
  title        = {Pith review of: Real-Time, Single-Ear, Wearable ECG Reconstruction, R-Peak Detection, and HR/HRV Monitoring},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SP4X73FC}},
  note         = {Machine review of arXiv:2505.01738}
}
read the original abstract

Biosignal monitoring, in particular heart activity through heart rate (HR) and heart rate variability (HRV) tracking, is vital in enabling continuous, non-invasive tracking of physiological and cognitive states. Recent studies have explored compact, head-worn devices for HR and HRV monitoring to improve usability and reduce stigma. However, this approach is challenged by the current reliance on wet electrodes, which limits usability, the weakness of ear-derived signals, making HR/HRV extraction more complex, and the incompatibility of current algorithms for embedded deployment. This work introduces a single-ear wearable system for real-time ECG (Electrocardiogram) parameter estimation, which directly runs on BioGAP, an energy-efficient device for biosignal acquisition and processing. By combining SoA in-ear electrode technology, an optimized DeepMF algorithm, and BioGAP, our proposed subject-independent approach allows for robust extraction of HR/HRV parameters directly on the device with just 36.7 uJ/inference at comparable performance with respect to the current state-of-the-art architecture, achieving 0.49 bpm and 25.82 ms for HR/HRV mean errors, respectively and an estimated battery life of 36h with a total system power consumption of 7.6 mW. Clinical relevance: The ability to reconstruct ECG signals and extract HR and HRV paves the way for continuous, unobtrusive cardiovascular monitoring with head-worn devices. In particular, the integration of cardiovascular measurements in everyday-use devices (such as earbuds) has potential in continuous at-home monitoring to enable early detection of cardiovascular irregularities.

Figures

Figures reproduced from arXiv: 2505.01738 by the authors.

Figure 1
Figure 1. Overview of the measurement setup (left) and algorithmic flow: (1) autoencoder training, (2) R-peak detection [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Rolling window R-peak refinement process. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Inference example illustrating the input signals for [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: Biopotential measurements aligned to the ground [PITH_FULL_IMAGE:figures/full_fig_p005_3.png]
Figure 7
Figure 7. Figure 7: Mean HR. and HRV error (all subjects, LOSO CV). [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 6
Figure 6. Figure 6: presents the achieved precision, recall, and F1-score. R-peak Precision, Recall, and F1 Score DeepMF Single Ear DeepMF Cross Ear Mini DeepMF Single Ear Mini DeepMF Cross Ear 0.7 0.75 0.8 0.85 0.9 0.95 1 Score Precision Recall F1 Score [PITH_FULL_IMAGE:figures/full_fig…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

  1. [1]

    Wearables and the medical revolution,

    J. Dunn et al., “Wearables and the medical revolution,” Personalized medicine, vol. 15, no. 5, pp. 429–448, 2018

    work page 2018

  2. [2]

    Advances in biosignal sensing and signal process- ing methods with wearable devices,

    J. Matthews et al., “Advances in biosignal sensing and signal process- ing methods with wearable devices,” Analysis & Sensing, vol. 3, no. 2, p. e202200062, 2023

    work page 2023

  3. [3]

    A comprehensive accuracy assessment of samsung smartwatch heart rate and heart rate variability,

    F. Sarhaddi et al., “A comprehensive accuracy assessment of samsung smartwatch heart rate and heart rate variability,” PloS one , vol. 17, no. 12, p. e0268361, 2022

    work page 2022

  4. [4]

    SoftPulse Products: Innovative Solutions for Wearable Sensors,

    Datwyler Group, “SoftPulse Products: Innovative Solutions for Wearable Sensors,” https://datwyler.com/company/innovation/ softpulse/products, 2025, accessed: 2025-01-25

    work page 2025

  5. [5]

    Gapses: Versatile smart glasses for comfortable and fully-dry acquisition and parallel ultra-low-power processing of eeg and eog,

    S. Frey et al. , “Gapses: Versatile smart glasses for comfortable and fully-dry acquisition and parallel ultra-low-power processing of eeg and eog,” IEEE TBioCAS, pp. 1–11, 2024

    work page 2024

  6. [6]

    Hearables: feasibility of recording cardiac rhythms from head and in-ear locations,

    W. von Rosenberg et al., “Hearables: feasibility of recording cardiac rhythms from head and in-ear locations,” Royal Society open science , vol. 4, no. 11, pp. 171 214–171 214, 2017

    work page 2017

  7. [7]

    Hearables: Feasibility and validation of in-ear electrocardiogram,

    G. Hammour et al. , “Hearables: Feasibility and validation of in-ear electrocardiogram,” in IEEE EMBC, 2019, pp. 5777–5780

    work page 2019

  8. [8]

    The deep-match framework: R-peak detection in ear-ecg,

    H. J. Davies et al., “The deep-match framework: R-peak detection in ear-ecg,” IEEE TBioCAS, vol. 71, no. 7, pp. 2014–2021, 2024

    work page 2014

Show all 16 references
  1. [9]

    A wireless system for eeg acquisition and processing in an earbud form factor with 600 hours battery lifetime,

    M. Guermandi et al. , “A wireless system for eeg acquisition and processing in an earbud form factor with 600 hours battery lifetime,” in Proc. IEEE EMBC , 2022, pp. 3139–3145

    2022

  2. [10]

    Hearables: feasibility of recording cardiac rhythms from single in-ear locations,

    M. Yarici et al., “Hearables: feasibility of recording cardiac rhythms from single in-ear locations,” Royal Society open science , vol. 11, no. 1, pp. 221 620–17, 2024

    2024

  3. [11]

    Enabling r-peak detection in wearable ecg: Combining matched filtering and hilbert transform,

    T. Chanwimalueang et al., “Enabling r-peak detection in wearable ecg: Combining matched filtering and hilbert transform,” in IEEE DSP . IEEE, 2015, pp. 134–138

    2015

  4. [12]

    Hearables: Heart rate variability from ear electrocar- diogram and ear photoplethysmogram (ear-ecg and ear-ppg),

    H. Tian et al., “Hearables: Heart rate variability from ear electrocar- diogram and ear photoplethysmogram (ear-ecg and ear-ppg),” in IEEE EMBC, vol. 2023. United States: IEEE, 2023, pp. 1–5

    2023

  5. [13]

    In-ear ecg signal enhancement with denoising convolutional autoencoders,

    E. Occhipinti et al. , “In-ear ecg signal enhancement with denoising convolutional autoencoders,” 2024. [Online]. Available: https://arxiv. org/abs/2409.05891

    2024 arXiv

  6. [14]

    Biogap: a 10-core fp-capable ultra-low power iot processor, with medical-grade afe and ble connectivity for wearable biosignal processing,

    S. Frey et al. , “Biogap: a 10-core fp-capable ultra-low power iot processor, with medical-grade afe and ble connectivity for wearable biosignal processing,” IEEE COINS, 2023

    2023

  7. [15]

    Nntool: Neural network tool for gap proces- sors,

    G. Technologies, “Nntool: Neural network tool for gap proces- sors,” https://greenwaves-technologies.com/tools-and-software/, ac- cessed: 2024-05-24

    2024

  8. [16]

    Hearables: In-Ear Multimodal Data Fusion for Robust Heart Rate Estimation,

    M. ˙Zyli´nski et al. , “Hearables: In-Ear Multimodal Data Fusion for Robust Heart Rate Estimation,” BioMedInformatics, vol. 4, no. 2, pp. 911–920, Apr. 2024

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.