Ambiguity-Resolved Micro-Doppler Construction for Asynchronous Bistatic Sensing

J. Andrew Zhang; Kai Wu; Yiyi Xu; Zhongqin Wang

REVIEW 4 major objections 4 minor 17 references

Ambiguity-Resolved Micro-Doppler Construction for Asynchronous Bistatic Sensing

T0 review · 4 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a delay-AoA-Doppler filtering pipeline can resolve mirror ambiguity and suppress residual second-order by-products left after differential CACC, producing clean micro-Doppler signatures for asynchronous bistatic WiFi

desk verdict Novel CACC-plus-DAD pipeline for asynchronous bistatic micro-Doppler, strong Widar3.0 results, but the Eq. (6) normalization reintroduces the denominator instability it warns about. read the letter →

arxiv 2607.20108 v1 pith:HRFNDJ52 submitted 2026-07-22 eess.SP

Yiyi Xu , Zhongqin Wang , Kai Wu , J. Andrew Zhang This is my paper · ORCID

classification eess.SP MSC 94A1294A29

keywords micro-Dopplerbistaticsensingcross-antennaconjugatemultiplicationISACgesturerecognitionasynchronousclocksdelay-AoA-Dopplertransform

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 is trying to establish that, after cross-antenna conjugate multiplication (CACC) and cyclic differencing remove the dominant mirror ambiguity, the leftover second-order by-products—which have distinct delay and angle-of-arrival (AoA) structure from the desired motion component—can be suppressed by a lightweight multidimensional filtering pipeline. If true, it would give WiFi-based gesture and activity recognition a physically interpretable, domain-robust micro-Doppler representation without requiring hardware phase synchronization. The authors test this on Widar3.0 and report strong cross-domain generalization, suggesting that explicit residual suppression is the key to clean and unambiguous bistatic sensing.

What carries the argument

The key mechanism is the delay-AoA-Doppler (DAD) transform-domain pipeline, applied after CACC and cyclic pairwise antenna differencing. It uses a DFT along the subcarrier dimension to map to delay bins, an M_a-point DFT along the antenna dimension to check for ordered steering structure, and a final DFT along slow time to extract Doppler. The residual second-order by-products are shown to be incoherent across these domains, so averaging over delay and AoA bins attenuates them while preserving the target's kinematic ridges.

What would settle it

A direct test would be to take a bistatic measurement with a known moving target and a controlled static background, then compare the output of this pipeline against the ground-truth micro-Doppler from a synchronized monostatic reference. If the target's true Doppler signature is not recovered when the static channel has a deep null at one subcarrier, the normalization assumption is falsified.

Watch

Extended reading notes

Core claim

The central claim is that the nonlinear term left after differential CACC—the second-order by-product X_m1 ⊙ X*_m2—does not share the same delay-Doppler-AoA structure as the desired motion-induced response, so it can be filtered out in a transform domain. The paper derives that the desired component focuses at physical delays r_p, AoA bin µ_p, and Doppler f^D_p, while the residual auto-terms collapse toward near-zero delay and Doppler, and the cross-terms appear at pairwise-difference delay coordinates without ordered AoA steering. Exploiting this, the proposed delay-AoA-Doppler (DAD) transform plus a variance-reducing aggregation produces a compact micro-Doppler spectrogram with higher per-

Load-bearing premise

The pipeline divides the gain-normalized CACC output by an estimated static term bU; if bU is small at any subcarrier due to a static-channel null or reference-antenna fading, that division amplifies noise and residual static mismatch, and the derived structure that the DAD filtering relies on breaks down.

Editorial extensions

If this is right

If correct, the proposed pipeline would enable asynchronous bistatic WiFi sensing without clock synchronization, since CACC removes random phase while preserving linear structure.
The reported gains over representative baselines suggest that residual second-order suppression is a major factor in cross-domain generalization for micro-Doppler features.
The approach could be extended to other ISAC scenarios or different antenna configurations, since it relies only on the presence of a multi-antenna receiver and the structure of motion-induced paths.
The aggregation step reduces noise variance and stabilizes Doppler signatures, making the feature more robust to environmental shifts, which is promising for real-world deployment.

Reading between the lines

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

The paper's framework suggests that any bistatic system with at least three antennas could use a similar cyclic-differencing plus transform-domain filtering approach, potentially extending to mmWave or sub-6 GHz ISAC links.
If the static-normalization step is the weakest link, a practical implementation might need to adaptively detect and skip subcarriers where the static estimate bU is near zero, or use a regularized inversion, to avoid noise amplification in those bins.
The claim that residual terms lack ordered AoA steering is specific to the cyclic differencing construction; a different differencing pattern might reduce the residuals further or change their structure, offering a testable extension.
The method's gain in cross-domain accuracy over BVP and MCP suggests that clean micro-Doppler ridges carry more transferable kinematic information than hand-crafted velocity profiles, which could inform future feature design for wireless sensing.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Dear colleague,

This is a credible engineering advance. The pipeline—CACC, cyclic antenna-pair differencing, then delay-AoA-Doppler filtering—is a genuinely new combination for cleaning up the residual second-order by-products that survive differential CACC. The ablation shows each stage contributes, and the Widar3.0 evaluation is substantial: 105k samples, leave-one-domain-out, with a performance gap over baselines large enough not to be written off as classifier luck. Credit where it's due.

The soft spots are real but not fatal. The stress-test note about Eq. (6) holds up on reading. After arguing that CACC avoids denominator-driven amplification, the paper divides by a per-subcarrier static estimate bU to produce W. If bU is near zero on any subcarrier—due to a static null or reference fading—noise is amplified by 1/|bU|^2, which is precisely the CSI-ratio instability they wanted to avoid. No conditioning analysis, no regularization, no threshold is reported. This is an unaddressed risk rather than a demonstrated failure, but it needs to be handled.

Second, the word "suppresses" oversells what the math shows. The delay transform redistributes residual by-products to difference delay coordinates, and the AoA transform makes them incoherent at target bins, but the final aggregation in Eq. (15) averages over all delay/AoA bins, re-including those residual magnitudes. What you get is variance reduction, not removal. That's a reasonable engineering heuristic, but the abstract should say so.

Third, the empirical section could be tightened: no error bars, one classifier for all baselines, and the baseline features are reformatted to a common input size, which may not match the original design. The margin is large enough that the main conclusion is probably robust, but the exact numbers should be treated with some caution.

Bottom line: this is a novel signal-processing pipeline for an important problem, backed by a public benchmark and honest ablations. The theory needs either softening or additional analysis, especially around Eq. (6). I'd send it to peer review; the issues are addressable in revision. Worth bringing to a reading group if you follow WiFi sensing or ISAC.

Referee Report

4 major / 4 minor

Summary. The paper proposes a signal-processing pipeline for constructing micro-Doppler spectrograms from asynchronous bistatic WiFi CSI. It starts from a stated signal model and uses cross-antenna conjugate multiplication (CACC) to remove the shared random phase, estimates and subtracts the static background, then normalizes by the static estimate. Cyclic antenna differencing suppresses the dominant mirror component, and a delay-AoA-Doppler transform is used to attenuate residual second-order by-products. The final representation aggregates delay-AoA-Doppler response magnitudes into a spectrogram. Experiments on Widar3.0 with a ResNet-50 classifier are reported, with accuracy between 90.1% and 98.2% across location, orientation, environment, and user shifts, outperforming three feature baselines. The paper claims that the framework resolves the mirror ambiguity and suppresses residual second-order by-products, and that the resulting representation generalizes across domains.

Significance. If the technical claims are correct, the paper offers a practical, physically interpretable Doppler feature for asynchronous bistatic ISAC without requiring hardware phase synchronization. The central derivation is not circular: the transform-domain filtering is derived from an explicit signal model and the downstream classifier is only used for evaluation. The paper also provides a formal perturbation comparison between CSI-ratio and CACC, a clear ablation protocol on a public benchmark, and reproducible-looking comparisons on Widar3.0. These are strengths. However, two load-bearing points—the static normalization in Eq. (6) and the final aggregation in Eq. (15)—need more careful support before the claimed suppression of residual by-products can be accepted.

major comments (4)

[§II-B, Eq. (6)] The pipeline defines W = Vhat ⊘ Uhat after Claim 1 argued that CACC avoids the denominator-driven instability of CSI-ratio. Uhat is an estimated per-subcarrier static CACC term obtained by time-averaging over a 0.128 s CPI. If any subcarrier lies near a static-channel null, or if the static component is weak relative to the dynamic component, |Uhat| can be small, and the division amplifies noise and static-mismatch terms by 1/|Uhat| (power 1/|Uhat|^2), exactly the mechanism attributed to CSI-ratio in Eq. (3). All subsequent stages in Eqs. (7)–(15) operate on W, so the claimed linear structure is premised on Uhat being well conditioned on every subcarrier. The paper reports no conditioning analysis, regularization, or threshold for when Uhat is too small. Furthermore, Uhat is a time average; a nonzero-mean dynamic component biases it, so the denominator can be inaccurate even where it is
[§III-D, Eq. (15)] The abstract and conclusion claim that residual second-order by-products are 'suppressed', but the final aggregation re-includes them. Eqs. (10)–(14) show that the residuals are redistributed to pairwise-difference delay and Doppler coordinates and to mismatched AoA bins; they are not zeroed. Eq. (15) sums the magnitudes over all NrMa delay–AoA bins, so residual energy at non-target bins contributes directly to the spectrogram. The statement before Eq. (15) that averaging reduces residual variance relies on the residuals being zero-mean, mutually uncorrelated fluctuations with common variance σ². These residuals are deterministic products of signal components, not stochastic fluctuations, and no evidence is given for the uncorrelated/common-variance assumption. Without quantifying the residual contribution at non-target bins or restricting the aggregation to target bins, the suppression
[§III-B, Eqs. (12)–(13)] The AoA-domain separation is the main mechanism for suppressing second-order residuals, but the key second-order spatial response is not given. The text states that its 'exact closed-form expression is omitted for space constraints' and provides only a qualitative description plus an example for M=3. Since the central claim is that residuals are not coherently focused at the target bins, the referee needs either the full expression or a numerical evaluation showing the leakage contribution as a function of M, Ma, and the AoA difference (µp − µq). As written, the key filtering property is not verifiable.
[§IV-A, dataset choice] The evaluation on Widar3.0 does not validate the asynchronous-bistatic component of the claim. The signal model in §II-A explicitly includes random phase offsets from transceiver clock asynchrony, but no evidence is provided that the Widar3.0 recordings contain such asynchrony. If the data is from conventional synchronous WiFi, the experiments do not test the central motivation of the paper. A synthetic asynchronous test, or at minimum a clear statement of the sync status of the dataset and a discussion of how the method performs without asynchrony, is needed.

minor comments (4)

[Eqs. (1), (4), (6)] The symbol W is used both for additive noise in Eq. (1) and for the static-normalized motion term in Eq. (6). This is confusing; a distinct symbol such as N_m or E_m for noise would improve readability.
[Eq. (10)] The delay-transform notation 'ZT_{N→Nr} F^T_{Nr}/N' is not fully defined; the dimensions of the zero-padding and DFT matrices should be stated explicitly so the expression is unambiguous.
[§III-A, Eq. (11)] The definition of q(r) uses the Dirichlet kernel D_N, but the argument −2π(r∆f/c + ℓ/Nr) mixes continuous and discrete quantities; a short derivation or reference would clarify the bin alignment.
[§IV-C] The cross-domain results in Fig. 4 are reported as bar charts without error bars or significance tests. Given the claims of about 18 percentage points improvement, the variability across random seeds should be reported.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the micro-Doppler construction is derived algebraically from a stated signal model and evaluated on an external dataset against external baselines.

full rationale

The paper's derivation chain is self-contained in the relevant sense. The CACC representation (Eq. 4–5), cyclic differencing (Eq. 7), and delay-AoA-Doppler transforms (Eqs. 10–14) are algebraic consequences of the stated signal model, not fitted to downstream labels. Claim 1 uses a first-order perturbation computation to compare CSI-ratio and CACC; whether one agrees with its conclusion, the comparison is not circular because each error expression is derived from the cited definitions. The subsequent normalization W = Vhat ./ bU reintroduces a denominator-sensitivity concern, but this is a numerical robustness issue, not a circularity: no parameter is fitted to the evaluation data and no quantity is defined in terms of the claimed prediction. The self-citations to prior CACC/tracking work are used as background and for the signal model; the novelty -- residual suppression via multidimensional transform-domain filtering -- is argued from the equations in the paper. The final evaluation uses the public Widar3.0 dataset and compares against independent baselines (BVP, MCP, CSI-ratio), so the empirical claim is externally grounded. Processing hyperparameters (CPI length, FFT sizes, AoA bins) are implementation choices, not fitted predictions. No step reduces by construction to its own input, and no load-bearing argument depends on an unverified self-citation. Therefore, no significant circularity is found.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central derivation rests on a series of stated approximations: a single-reference static model, frequency-flat scattering gains, quasi-static reflection coefficients, separability of AoA and slow-time coefficients, and zero-mean residual magnitude fluctuations. The normalization-by-static-estimate step reintroduces a denominator risk that the paper itself associates with CSI-ratio. No new physical entities are introduced.

free parameters (5)

Coherent processing interval (CPI) = 0.128 s
Chosen in §IV-A; sets Doppler resolution and the quasi-static assumption.
Delay/range window length N_r = 32 m window, 1 m resolution (N_r=32)
Zero-padded DFT length in §III-A/§IV-A; affects delay focusing and cost.
AoA bin count M_a = 8
Spatial DFT size in §III-B/§IV-A; controls angular resolution of the steering filter.
Temporal interpolation shape = 64×128
Micro-Doppler matrix size after interpolation in §IV-A; chosen to match ResNet input.
AGC gain compensation proxy g-hat = not specified
Derived from 'reported signal strength' in §II-B but no formula is given; affects static-term estimation.

assumptions (6)

domain assumption The bistatic CSI follows H_m = (A_m ⊙ Φ_m) ⊙ (S_m + X_m) + W_m with multiplicative AGC, phase-offset, static and dynamic components.
Eq. (1)–(2); this model is the foundation of the entire derivation.
ad hoc to paper The aggregate static background is locally approximated by a single dominant equivalent reference component (θ_s, d_s, ρ_{m,s}).
Eq. (2); this simplification makes the static-normalization in Eq. (6) tractable but is not externally validated.
domain assumption Effective scattering gains are approximately frequency-flat within the signal bandwidth.
Invoked before Eq. (10) to justify the delay-domain representation; WiFi channels can be frequency selective.
domain assumption Reflection coefficients are quasi-static within a CPI: ρ̃_p ≈ ρ̃_p 1_K.
Invoked before Eq. (14) for the Doppler transform; human motion may violate this within 0.128 s.
domain assumption The antenna-dependent coefficient is separable: γ_m(θ_p)/γ_m(θ_s) ρ̃_{m,p} ≈ e^{-j(m-1)μ_p} ρ̃_p.
Invoked before Eq. (12) to obtain the ordered AoA steering structure.
ad hoc to paper Residual magnitude fluctuations around their local means are zero-mean, mutually uncorrelated, with common variance σ².
Invoked in §III-D to claim SNR improvement from aggregation; no empirical justification is given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Ambiguity-Resolved Micro-Doppler Construction for Asynchronous Bistatic Sensing." pith.science (2026). https://pith.science/paper/HRFNDJ52

@misc{pith2026260720108,
  author       = {Pith},
  title        = {Pith review of: Ambiguity-Resolved Micro-Doppler Construction for Asynchronous Bistatic Sensing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HRFNDJ52}},
  note         = {Machine review of arXiv:2607.20108}
}

read the original abstract

Integrated sensing and communications (ISAC) can turn wireless networks into pervasive sensing platforms, but bistatic deployments are impaired by transceiver clock asynchrony, which induces random phase fluctuations in channel state information (CSI). CSI-ratio sanitization introduces nonlinear distortion that limits multi-target scalability and complicates delay- and angle-of-arrival (AoA)-domain processing. Cross-antenna conjugate multiplication (CACC) preserves a linear structure, but leaves mirror ambiguity and second-order by-products that corrupt motion-induced Doppler signatures. We develop a micro-Doppler construction framework that resolves the ambiguity and suppresses these residual by-products. Cyclic differencing first attenuates the dominant mirror component. We then exploit the facts that residual terms occur at differenced delay coordinates and lack an ordered AoA steering structure, designing a lightweight delay-AoA-Doppler pipeline that isolates the desired kinematic response without coherently accumulating residuals at target bins. The filtered responses are aggregated into a higher-SNR micro-Doppler representation. Ablation studies confirm the value of residual suppression and multidimensional filtering. On a large-scale WiFi gesture dataset, the proposed representation generalizes across four domain factors, attaining 90.1%-98.2% accuracy and outperforming representative baselines by about 18 percentage points on average.

Figures

Figures reproduced from arXiv: 2607.20108 by the authors.

**Figure 2.** Impact of artifact suppression on micro-Doppler signatures. (a)-(c) show “Push & Pull” and (d)-(f) show “Slide”. (a) [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Confusion matrices for in-domain gesture recognition [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Accuracy comparison between the proposed micro-Doppler representation and three baselines across domains. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 1 linked inside Pith

[1]

Guest editorial: Integrated sensing and communications for 6g,

C. Masouros, J. A. Zhang, F. Liu, L. Zheng, H. Wymeersch, and M. Di Renzo, “Guest editorial: Integrated sensing and communications for 6g,”IEEE Wireless Communications, vol. 30, no. 1, pp. 14–15, 2023

2023
[2]

Isac: From human to environmental sensing,

K. Wu, Z. Wang, S.-L. Chen, J. A. Zhang, and Y . J. Guo, “Isac: From human to environmental sensing,”IEEE Journal of Selected Topics in Electromagnetics, Antennas and Propagation, vol. 1, no. 1, pp. 84–98, 2025

2025
[3]

Towards SISO bistatic sensing for ISAC,

Z. Wang, J. A. Zhang, K. Wu, M. Xu, and Y . J. Guo, “Towards SISO bistatic sensing for ISAC,”arXiv preprint arXiv:2508.12614, 2025

arXiv 2025
[4]

Sensing in bistatic isac systems with clock asynchronism: A signal processing perspective,

K. Wu, J. Pegoraro, F. Meneghello, J. A. Zhang, J. O. Lacruz, J. Widmer et al., “Sensing in bistatic isac systems with clock asynchronism: A signal processing perspective,”IEEE Signal Processing Magazine, vol. 41, no. 5, pp. 31–43, 2024

2024
[5]

Csi-ratio-based doppler frequency estimation in integrated sensing and communications,

X. Li, J. A. Zhang, K. Wu, Y . Cui, and X. Jing, “Csi-ratio-based doppler frequency estimation in integrated sensing and communications,”IEEE Sensors Journal, vol. 22, no. 21, pp. 20 886–20 895, 2022

2022
[6]

Passive human tracking with wifi point clouds,

Z. Wang, J. A. Zhang, H. Zhang, M. Xu, and J. Guo, “Passive human tracking with wifi point clouds,”IEEE Internet of Things Journal, vol. 12, no. 5, pp. 5528–5543, 2025

2025
[7]

Widar3.0: Zero-effort cross-domain gesture recognition with wi-fi,

Y . Zhang, Y . Zheng, K. Qian, G. Zhang, Y . Liu, C. Wuet al., “Widar3.0: Zero-effort cross-domain gesture recognition with wi-fi,”IEEE Trans- actions on Pattern Analysis and Machine Intelligence, vol. 44, no. 11, pp. 8671–8688, 2022

2022
[8]

Wihf: Gesture and user recognition with wifi,

C. Li, M. Liu, and Z. Cao, “Wihf: Gesture and user recognition with wifi,”IEEE Transactions on Mobile Computing, vol. 21, no. 2, pp. 757– 768, 2022

2022

Show all 17 references

[9]

Toward robust device-free gesture recognition based on intrinsic spectrogram of mmwave signals,

J. Wu, J. Wang, Q. Gao, M. Cheng, M. Pan, and H. Zhang, “Toward robust device-free gesture recognition based on intrinsic spectrogram of mmwave signals,”IEEE Internet of Things Journal, vol. 9, no. 19, pp. 19 318–19 329, 2022

2022
[10]

Handgest: Hierarchical sensing for robust-in-the-air handwriting recognition with commodity wifi devices,

J. Zhang, Y . Li, H. Xiong, D. Dou, C. Miao, and D. Zhang, “Handgest: Hierarchical sensing for robust-in-the-air handwriting recognition with commodity wifi devices,”IEEE Internet of Things Journal, vol. 9, no. 19, pp. 19 529–19 544, 2022

2022
[11]

Time series focused neural network for accurate wireless human gesture recognition,

H. Huang, L. Lin, L. Zhao, S. Ding, and H. Huang, “Time series focused neural network for accurate wireless human gesture recognition,”IEEE Transactions on Network Science and Engineering, vol. 13, pp. 118–129, 2026

2026
[12]

Real-time cross- domain gesture and user identification via cots wifi,

C. Cao, Y . Ding, M. Dai, W. Gong, and X. Zhao, “Real-time cross- domain gesture and user identification via cots wifi,”IEEE Transactions on Mobile Computing, vol. 24, no. 6, pp. 5124–5137, 2025

2025
[13]

Crossfi: A cross domain wi-fi sensing framework based on siamese network,

Z. Zhao, T. Chen, Z. Cai, X. Li, H. Li, Q. Chenet al., “Crossfi: A cross domain wi-fi sensing framework based on siamese network,”IEEE Internet of Things Journal, vol. 12, no. 12, pp. 20 138–20 155, 2025

2025
[14]

V . C. Chen,The micro-Doppler effect in radar. Artech house, 2019

2019
[15]

Single-target real-time passive wifi tracking,

Z. Wang, J. A. Zhang, M. Xu, and Y . J. Guo, “Single-target real-time passive wifi tracking,”IEEE Transactions on Mobile Computing, vol. 22, no. 6, pp. 3724–3742, 2023

2023
[16]

Rethink- ing doppler effect for accurate velocity estimation with commodity wifi devices,

K. Niu, X. Wang, F. Zhang, R. Zheng, Z. Yao, and D. Zhang, “Rethink- ing doppler effect for accurate velocity estimation with commodity wifi devices,”IEEE Journal on Selected Areas in Communications, vol. 40, no. 7, pp. 2164–2178, 2022

2022
[17]

Deep residual learning for image recognition,

K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

2016

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Guest editorial: Integrated sensing and communications for 6g,

C. Masouros, J. A. Zhang, F. Liu, L. Zheng, H. Wymeersch, and M. Di Renzo, “Guest editorial: Integrated sensing and communications for 6g,”IEEE Wireless Communications, vol. 30, no. 1, pp. 14–15, 2023

2023

[2] [2]

Isac: From human to environmental sensing,

K. Wu, Z. Wang, S.-L. Chen, J. A. Zhang, and Y . J. Guo, “Isac: From human to environmental sensing,”IEEE Journal of Selected Topics in Electromagnetics, Antennas and Propagation, vol. 1, no. 1, pp. 84–98, 2025

2025

[3] [3]

Towards SISO bistatic sensing for ISAC,

Z. Wang, J. A. Zhang, K. Wu, M. Xu, and Y . J. Guo, “Towards SISO bistatic sensing for ISAC,”arXiv preprint arXiv:2508.12614, 2025

arXiv 2025

[4] [4]

Sensing in bistatic isac systems with clock asynchronism: A signal processing perspective,

K. Wu, J. Pegoraro, F. Meneghello, J. A. Zhang, J. O. Lacruz, J. Widmer et al., “Sensing in bistatic isac systems with clock asynchronism: A signal processing perspective,”IEEE Signal Processing Magazine, vol. 41, no. 5, pp. 31–43, 2024

2024

[5] [5]

Csi-ratio-based doppler frequency estimation in integrated sensing and communications,

X. Li, J. A. Zhang, K. Wu, Y . Cui, and X. Jing, “Csi-ratio-based doppler frequency estimation in integrated sensing and communications,”IEEE Sensors Journal, vol. 22, no. 21, pp. 20 886–20 895, 2022

2022

[6] [6]

Passive human tracking with wifi point clouds,

Z. Wang, J. A. Zhang, H. Zhang, M. Xu, and J. Guo, “Passive human tracking with wifi point clouds,”IEEE Internet of Things Journal, vol. 12, no. 5, pp. 5528–5543, 2025

2025

[7] [7]

Widar3.0: Zero-effort cross-domain gesture recognition with wi-fi,

Y . Zhang, Y . Zheng, K. Qian, G. Zhang, Y . Liu, C. Wuet al., “Widar3.0: Zero-effort cross-domain gesture recognition with wi-fi,”IEEE Trans- actions on Pattern Analysis and Machine Intelligence, vol. 44, no. 11, pp. 8671–8688, 2022

2022

[8] [8]

Wihf: Gesture and user recognition with wifi,

C. Li, M. Liu, and Z. Cao, “Wihf: Gesture and user recognition with wifi,”IEEE Transactions on Mobile Computing, vol. 21, no. 2, pp. 757– 768, 2022

2022