Pith. sign in

REVIEW 2 major objections 1 minor 24 references

Human Walking Sensing and Pose Estimation in the 6 GHz Band Using Amplitude and Phase CSI

T0 review · 2 major / 1 minor · reviewed 2026-06-27 · grok-4.3

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

Pith's one-line read Reliable human pose reconstruction is possible from 6 GHz OFDM CSI using adapted deep learning models.

desk verdict The paper adapts four existing pose models to 6 GHz OFDM CSI, shows amplitude often suffices, and reports workable numbers on an open dataset, but provides almost no evaluation details on splits or generalization. read the letter →

arxiv 2606.10048 v1 pith:626NQCI4 submitted 2026-06-08 eess.SP

Zhaorui Yin , Mattia Brambilla , Monica Nicoli This is my paper · ORCID
classification eess.SP
keywords humanposeestimationchannelstateinformation6GHzbandOFDMdeeplearningwirelesssensingamplitudephaseCSI
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

The paper establishes that human walking poses can be estimated from amplitude and phase of channel state information collected over multiple radio links in the 6 GHz band. Four existing deep learning models are adapted to the OFDM CSI structure and tested on an open-access dataset using standard pose metrics. A sympathetic reader would care if true because the method uses ordinary wireless signals rather than cameras or wearables, enabling sensing in environments where visual monitoring is impractical or undesirable. Results show DT-Pose performs best overall, amplitude alone often matches combined amplitude-phase input, and phase adds value mainly as a complement.

What carries the argument

Adapted deep learning architectures (DT-Pose, MetaFi++, HPE-Li, VST-Pose) that process amplitude and phase of multistatic 6 GHz OFDM CSI to output 3D body joint positions.

What would settle it

Repeating the evaluation on a new dataset collected with different subjects, room layouts, or walking patterns and obtaining substantially higher PA-MPJPE and BLL values than those reported would falsify the reliability claim.

Watch

Extended reading notes

Core claim

Reliable human pose reconstruction can be achieved from 6 GHz OFDM CSI measurements, with DT-Pose providing the best overall accuracy. On average, amplitude-only CSI yields performance comparable to joint amplitude-phase processing, whereas phase information is more beneficial as a complementary feature rather than as a standalone input.

Load-bearing premise

The open-access dataset used for evaluation is representative of real-world indoor multistatic walking scenarios and the adapted deep learning models generalize to unseen subjects and environments beyond the specific training conditions.

Editorial extensions

If this is right

  • DT-Pose yields the lowest errors among the four adapted models on the tested metrics.
  • Amplitude-only CSI produces accuracy comparable to joint amplitude-phase processing on average.
  • Phase information improves results primarily when combined with amplitude rather than used alone.
  • The pipeline works for human walking inside the coverage area of an indoor multistatic 6 GHz network.

Reading between the lines

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

  • The method could be integrated into existing 6 GHz Wi-Fi infrastructure for contactless activity monitoring without dedicated hardware.
  • Performance differences between amplitude and phase inputs might change if the models were retrained on data with stronger multipath or higher mobility.
  • The approach might extend to related tasks such as gait analysis or fall detection by reusing the same CSI input features.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper investigates human pose estimation from 6 GHz OFDM CSI in an indoor multistatic network. It adapts four existing deep learning models (DT-Pose, MetaFi++, HPE-Li, VST-Pose) to jointly process amplitude and phase CSI for estimating poses of walking humans and evaluates performance on an open-access dataset using PA-MPJPE and BLL metrics. The central claims are that reliable reconstruction is achievable, DT-Pose yields the best accuracy, amplitude-only CSI performs comparably to joint amplitude-phase on average, and phase is useful mainly as a complementary feature.

Significance. If the empirical results hold under proper validation, the work would contribute to RF-based human sensing by showing feasibility of pose estimation at 6 GHz using standard CSI measurements and off-the-shelf models. The use of an open-access dataset and standard metrics (PA-MPJPE, BLL) supports reproducibility and allows direct comparison with vision-based methods. No machine-checked proofs or parameter-free derivations are present, but the multistatic 6 GHz focus and amplitude-phase ablation are relevant to the eess.SP community.

major comments (2)
  1. [Abstract and evaluation section] Abstract and performance evaluation section: no details are provided on the number of subjects, environment diversity, train/test split strategy (cross-subject vs. intra-subject or random), training procedures, cross-validation, or error bars on the reported PA-MPJPE/BLL values. These omissions are load-bearing for the claims of 'reliable' reconstruction and generalization beyond the specific dataset, as the skeptic note correctly identifies.
  2. [Model adaptation and results sections] Model adaptation and results sections: the extensions of DT-Pose, MetaFi++, HPE-Li, and VST-Pose to OFDM CSI amplitude/phase are described at a high level without specifying input tensor shapes, preprocessing steps for phase unwrapping or calibration, or hyperparameter choices. This prevents assessment of whether the reported superiority of DT-Pose and the amplitude-only vs. joint processing comparison are robust.
minor comments (1)
  1. [Methods] Notation for CSI matrices and the precise definition of the multistatic links should be clarified with an equation or diagram early in the methods.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We agree that additional details are needed to support the claims of reliable reconstruction and to enable reproducibility. The revised manuscript will incorporate the requested information on experimental setup and model adaptations.

read point-by-point responses

  1. Referee: [Abstract and evaluation section] Abstract and performance evaluation section: no details are provided on the number of subjects, environment diversity, train/test split strategy (cross-subject vs. intra-subject or random), training procedures, cross-validation, or error bars on the reported PA-MPJPE/BLL values. These omissions are load-bearing for the claims of 'reliable' reconstruction and generalization beyond the specific dataset, as the skeptic note correctly identifies.

    Authors: We acknowledge these omissions in the current version and agree they are important for evaluating generalization. In the revised manuscript, the evaluation section will be expanded to specify the number of subjects, environment diversity, the train/test split strategy (cross-subject), training procedures, cross-validation approach, and error bars or standard deviations on the PA-MPJPE and BLL metrics. revision: yes

  2. Referee: [Model adaptation and results sections] Model adaptation and results sections: the extensions of DT-Pose, MetaFi++, HPE-Li, and VST-Pose to OFDM CSI amplitude/phase are described at a high level without specifying input tensor shapes, preprocessing steps for phase unwrapping or calibration, or hyperparameter choices. This prevents assessment of whether the reported superiority of DT-Pose and the amplitude-only vs. joint processing comparison are robust.

    Authors: We agree that the model adaptation descriptions are currently high-level and that more specifics are required. The revised manuscript will detail the input tensor shapes, preprocessing steps for phase unwrapping and calibration, and hyperparameter choices for each adapted model. This will allow better assessment of the DT-Pose superiority and amplitude-phase comparisons. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: empirical evaluation on external dataset with adapted literature models

full rationale

The paper adapts four existing deep learning architectures (DT-Pose, MetaFi++, HPE-Li, VST-Pose) from the literature to OFDM CSI inputs and reports performance on an open-access external dataset using standard metrics (PA-MPJPE, BLL). No mathematical derivation chain, fitted-parameter predictions, self-definitional steps, or load-bearing self-citations are present in the abstract or described pipeline. Claims rest on independent experimental results rather than reducing to the paper's own inputs by construction. Generalization assumptions are empirical limitations, not circularity.

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

This is an empirical machine learning application study adapting existing architectures to new signal data; no new free parameters, axioms, or invented entities are introduced beyond standard deep learning practices and the cited dataset.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Human Walking Sensing and Pose Estimation in the 6 GHz Band Using Amplitude and Phase CSI." pith.science (2026). https://pith.science/paper/626NQCI4

@misc{pith2026260610048,
  author       = {Pith},
  title        = {Pith review of: Human Walking Sensing and Pose Estimation in the 6 GHz Band Using Amplitude and Phase CSI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/626NQCI4}},
  note         = {Machine review of arXiv:2606.10048}
}
read the original abstract

This paper investigates human pose estimation from Orthogonal Frequency-Division Multiplexing (OFDM) signals in an indoor multistatic wireless network operating in the 6 GHz band. We design and validate a processing pipeline that exploits both the amplitude and phase of the Channel State Information (CSI) from multiple radio links to estimate the human body pose. Four deep learning architectures from the literature, namely DT-Pose, MetaFi++, HPE-Li, and VST-Pose, are adapted to the OFDM CSI structure and extended to jointly exploit the amplitude and phase information. The models estimate the pose of a human walking within the network coverage area. Performance evaluation is conducted on an open-access dataset using standard pose-estimation metrics such as Procrustes-aligned Mean Per-Joint Position Error (PA-MPJPE) and Bone Length Loss (BLL). Results indicate that reliable human pose reconstruction can be achieved from 6 GHz OFDM CSI measurements, with DT-Pose providing the best overall accuracy. On average, amplitude-only CSI yields performance comparable to joint amplitude-phase processing, whereas phase information is more beneficial as a complementary feature rather than as a standalone input.

Figures

Figures reproduced from arXiv: 2606.10048 by the authors.

Figure 2
Figure 2. CSI human pose estimation framework: three software defined radios [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 1
Figure 1. Frame structure and processing of the received OFDM signal for CSI [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. b and Fig. 3c. The PDP is computed for delay τ as: Pℓ,t(τ ) = 1 Ns X Ns i=1 [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Validation loss, PA-MPJPE and BLL metrics over [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: (a) Human pose reconstruction of DT-Pose after Procrustes alignment [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 2 canonical work pages

  1. [1]

    Non-invasive detection of moving and stationary human with WiFi,

    C. Wuet al., “Non-invasive detection of moving and stationary human with WiFi,”IEEE J. Sel. Areas Commun., vol. 33, no. 11, pp. 2329– 2342, 2015

    2015

  2. [2]

    Device-free radio vision for assisted living: Leveraging wireless channel quality information for human sensing,

    S. Savazziet al., “Device-free radio vision for assisted living: Leveraging wireless channel quality information for human sensing,”IEEE Signal Process. Mag., vol. 33, no. 2, pp. 45–58, 2016

    2016

  3. [3]

    A Bayesian approach to device-free localization: Modeling and experimental assessment,

    S. Savazzi, M. Nicoli, F. Carminati, and M. Riva, “A Bayesian approach to device-free localization: Modeling and experimental assessment,” IEEE J. Sel. Topics Signal Process., vol. 8, no. 1, pp. 16–29, 2014

    2014

  4. [4]

    A survey on CSI-based Wi-Fi sensing datasets and models with a focus on reproducibility,

    I. Guarino, D. Carra, M. Cominelli, F. Gringoli, and R. L. Cigno, “A survey on CSI-based Wi-Fi sensing datasets and models with a focus on reproducibility,”Comput. Commun., p. 108431, 2026

    2026

  5. [5]

    See through walls with WiFi!

    F. Adib and D. Katabi, “See through walls with WiFi!” inProc. ACM SIGCOMM, 2013, pp. 75–86

    2013

  6. [6]

    Capturing the human figure through a wall,

    F. Adib, C.-Y . Hsu, H. Mao, D. Katabi, and F. Durand, “Capturing the human figure through a wall,”ACM Trans. Graph., vol. 34, no. 6, pp. 1–13, 2015

    2015

  7. [7]

    WiSPE: A COTS Wi-Fi- based 2-D static human pose estimation,

    M. Xu, Z. Guo, L. Gui, B. Sheng, and F. Xiao, “WiSPE: A COTS Wi-Fi- based 2-D static human pose estimation,”IEEE Syst. J., vol. 17, no. 3, pp. 3560–3571, 2023

    2023

  8. [8]

    Target classification for integrated sensing and communication in industrial deployments,

    L. Barbieriet al., “Target classification for integrated sensing and communication in industrial deployments,” inProc. IEEE Int. Symp. Joint Commun. Sens. (JC&S), 2026, pp. 1–6

    2026

Show all 24 references
  1. [9]

    Learning-based analysis of 5G and WiFi CSI for indoor localization: Feature stability, model generalization, and performance trade-offs,

    Y . Ruanet al., “Learning-based analysis of 5G and WiFi CSI for indoor localization: Feature stability, model generalization, and performance trade-offs,”Neurocomputing, p. 133114, 2026

    2026

  2. [10]

    From point to space: 3D moving human pose estimation using commodity WiFi,

    Y . Wanget al., “From point to space: 3D moving human pose estimation using commodity WiFi,”IEEE Commun. Lett., vol. 25, no. 7, pp. 2235– 2239, 2021

    2021

  3. [11]

    Exposing the CSI: A systematic investigation of CSI-based Wi-Fi sensing capabilities and limitations,

    M. Cominelli, F. Gringoli, and F. Restuccia, “Exposing the CSI: A systematic investigation of CSI-based Wi-Fi sensing capabilities and limitations,” inProc. IEEE PerCom, 2023, pp. 81–90

    2023

  4. [12]

    MetaFi++: WiFi-enabled transformer-based human pose estimation for metaverse avatar simulation,

    Y . Zhouet al., “MetaFi++: WiFi-enabled transformer-based human pose estimation for metaverse avatar simulation,”IEEE Internet Things J., vol. 10, no. 16, pp. 14 128–14 136, 2023

    2023

  5. [13]

    HPE-Li: WiFi-enabled lightweight dual selective kernel convolution for human pose estimation,

    T. D. Gian, T. Dac Lai, T. Van Luong, K.-S. Wong, and V .-D. Nguyen, “HPE-Li: WiFi-enabled lightweight dual selective kernel convolution for human pose estimation,” inProc. ECCV 2024, A. Leonardiset al., Eds. Cham: Springer Nature Switzerland, 2025, pp. 93–111

    2024

  6. [14]

    Towards robust and realistic human pose estimation via WiFi signals,

    Y . Chen, J. Guo, S. Guo, J. Zhou, and D. Tao, “Towards robust and realistic human pose estimation via WiFi signals,”arXiv preprint arXiv:2501.09411, 2025

    2025 arXiv

  7. [15]

    VST-Pose: A velocity-integrated spatiotem-poral attention network for human WiFi pose estimation,

    X. Zhanget al., “VST-Pose: A velocity-integrated spatiotem-poral attention network for human WiFi pose estimation,”arXiv preprint arXiv:2507.09672, 2025

    2025

  8. [16]

    1–767, 2021

    “IEEE Standard for Information Technology–Telecommunications and Information Exchange between Systems Local and Metropolitan Area Networks–Specific Requirements Part 11: Wireless LAN Medium Ac- cess Control (MAC) and Physical Layer (PHY) Specifications Amend- ment 1: Enhanceme...

    2021

  9. [17]

    Feasibility Study on 6 GHz for LTE and NR in Licensed and Unlicensed Operations,

    3GPP, “Feasibility Study on 6 GHz for LTE and NR in Licensed and Unlicensed Operations,” 2024, TS 37.890 v19.0.0

    2024

  10. [18]

    Coexistence analysis of Wi-Fi 6E and 5G NR-U in the 6 GHz band,

    N. Keshtiarast and M. Petrova, “Coexistence analysis of Wi-Fi 6E and 5G NR-U in the 6 GHz band,” inProc. ICNS3. Association for Computing Machinery, 2025, p. 38–45

    2025

  11. [19]

    Next generation Wi-Fi and 5G NR-U in the 6 GHz bands: Opportunities and challenges,

    G. Naik, J.-M. Park, J. Ashdown, and W. Lehr, “Next generation Wi-Fi and 5G NR-U in the 6 GHz bands: Opportunities and challenges,”IEEE Access, vol. 8, pp. 153 027–153 056, 2020

    2020

  12. [20]

    Multi-static OFDM radar dataset for human activity analysis: Three USRP X410 based measurements with motion capture ground truth,

    B. Yanet al., “Multi-static OFDM radar dataset for human activity analysis: Three USRP X410 based measurements with motion capture ground truth,” 2025

    2025

  13. [21]

    Person-in-WiFi: Fine-grained person perception using WiFi,

    F. Wang, S. Zhou, S. Panev, J. Han, and D. Huang, “Person-in-WiFi: Fine-grained person perception using WiFi,” inProc. IEEE/CVF ICCV, 2019, pp. 5451–5460

    2019

  14. [22]

    PhaseFi: Phase fingerprinting for indoor localization with a deep learning approach,

    X. Wang, L. Gao, and S. Mao, “PhaseFi: Phase fingerprinting for indoor localization with a deep learning approach,” inProc. IEEE GLOBECOM, 2015, pp. 1–6

    2015

  15. [23]

    Error bounds of projection models in weakly supervised 3D human pose estimation,

    N. Klug, M. Einfalt, S. Brehm, and R. Lienhart, “Error bounds of projection models in weakly supervised 3D human pose estimation,” inProc. 3DV, 2020, pp. 898–907

    2020

  16. [24]

    Compositional human pose regression,

    X. Sun, J. Shang, S. Liang, and Y . Wei, “Compositional human pose regression,” inProc. IEEE/CVF ICCV, 2017, pp. 2621–2630

    2017

Pith tools

Reviewed June 27, 2026 · model on record in the stance chip above.