Low-Complexity Gridless Single-Snapshot DoA Estimation via Truncated Hankel Newton-MUSIC

Khaled B. Letaief; Ruoxiao Cao; Wentao Yu; Yi Gong

REVIEW 2 major objections 5 minor 15 references

Low-Complexity Gridless Single-Snapshot DoA Estimation via Truncated Hankel Newton-MUSIC

T0 review · 2 major / 5 minor · reviewed 2026-07-10 · grok-4.5

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

Pith's one-line read A fixed-size Hankel window plus Newton refinement makes single-snapshot DoA estimation linear in array size, with accuracy close to full square Hankel MUSIC.

desk verdict Solid engineering trade-off paper: fixed-row Hankel + Newton gives near-square accuracy at linear cost for single-snapshot DoA, demonstrated cleanly but only inside a comfortable regime. read the letter →

arxiv 2607.08108 v1 pith:5SY445HQ submitted 2026-07-09 eess.SP

Ruoxiao Cao , Wentao Yu , Yi Gong , Khaled B. Letaief This is my paper · ORCID

classification eess.SP

keywords single-snapshotDoAHankelspatialsmoothingtruncatedNewton-MUSICreconfigurableantennasISAClow-complexitysubspacemethods

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

In reconfigurable-antenna ISAC systems that must estimate directions of arrival from one snapshot, classical subspace methods fail because the sample covariance is rank-deficient, while full Hankel spatial smoothing restores rank at cubic cost in the number of ports. This paper shows that a deliberately truncated Hankel matrix whose row dimension is held fixed and independent of array size still recovers a usable signal subspace, so that correlation construction and SVD become linear in the number of antennas. Coarse grid peaks are then polished by a few second-order Newton steps in continuous angle, removing the quantization floor that dense-grid MUSIC leaves behind. Simulations with hundreds of ports confirm that the accuracy stays within roughly 1.5 dB of the expensive square-Hankel Newton baseline while runtime drops by more than two orders of magnitude at N=1024. The result matters for real-time sensing: large reconfigurable arrays can keep high angular resolution without waiting for multiple snapshots or paying cubic compute.

What carries the argument

Truncated Hankel matrix of fixed row dimension LN≪N: the single snapshot is rearranged into a rectangular Hankel matrix whose small row size keeps the smoothed correlation matrix tiny, while the long column dimension still supplies enough overlapping subarrays for subspace recovery; subsequent Newton updates on the continuous MUSIC cost then refine the coarse peaks.

What would settle it

Increase the number of sources or shrink the minimum angular separation while keeping LN fixed at the paper’s value (20); if the RMSE gap to square Hankel Newton-MUSIC then opens well beyond 1.5 dB or the method fails to resolve the sources, the truncation premise fails.

Watch

Extended reading notes

Core claim

When the Hankel row dimension is fixed at a small constant LN independent of array size N (and only slightly larger than the number of sources), Truncated Hankel Newton-MUSIC recovers a usable noise subspace, yields DoA RMSE within about 1.5 dB of full square Hankel Newton-MUSIC, and reduces the dominant complexity from O(N^{3}) to O(LN^{2} N), producing more than two orders of magnitude runtime reduction at N=1024.

Load-bearing premise

A fixed, small truncation length still leaves enough spatial diversity that the noise subspace of the tiny correlation matrix remains usable for MUSIC, and that any resolution loss is fully recovered by local Newton steps started from a coarse half-degree grid.

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

The one thing worth knowing is that they deliberately fix the Hankel row length LN independent of N, keep the long column for smoothing, and then Newton-refine coarse peaks. That is the actual move. Complexity then drops to O(LN^{2}N) instead of O(N^{3}), and at N=1024 they show more than two orders of magnitude runtime cut while RMSE stays within ~1.5 dB of the square-Hankel Newton baseline.

What is new is modest but real: the combination of fixed-row truncation (LN=20 for K=4) with second-order continuum refinement for the single-snapshot reconfigurable-array setting. Hankel smoothing and Newtonized continuum search already exist; the paper’s contribution is the deliberate truncation-plus-Newton package plus the matching complexity table and runtime curves. The math is standard and clean—SVD equivalence note, analytic gradient/Hessian of the MUSIC cost, Algorithm 1. Monte-Carlo design (M=2000, N=256, K=4, min sep 9°) is transparent and the four-way ablation (square/truncated × grid/Newton) isolates the two design choices. Fig. 1 and Fig. 2 support the accuracy–complexity claim under those conditions. Citations to Liao–Fannjiang, Mamandipoor, and classical MUSIC/ESPRIT are appropriate; self-cites are minor.

Soft spots are real but proportionate. LN is an empirical free parameter; there is no bound relating admissible K or minimum separation to LN, and the simulations never stress the regime where K approaches LN or sources sit closer than 9°. Newton cannot invent peaks the coarse truncated spectrum never produced. K is assumed known, no real RF data, no modern sparse-recovery baselines. Those are the usual systems-paper gaps, not load-bearing contradictions.

This is for people who actually need low-latency single-snapshot DoA on large reconfigurable arrays in ISAC. A serious referee should see it; the core claim is reproducible from the given equations and figures. I would engage if I were working the same latency–N trade-off; otherwise it is a clean incremental systems note.

Referee Report

2 major / 5 minor

Summary. The paper proposes Truncated Hankel Newton-MUSIC for single-snapshot DoA estimation with large reconfigurable arrays in ISAC. A single snapshot is rearranged into a rectangular Hankel matrix with fixed small row dimension LN ≪ N (LN+1 > K), yielding a low-dimensional smoothed correlation matrix whose SVD recovers a usable noise subspace at cost O(LN^{2}N + LN^{3}). Coarse-grid minima of the MUSIC cost J(θ) are then refined by second-order Newton updates that use analytic first and second derivatives of the truncated steering vector. Complexity analysis (Table I) and Monte-Carlo simulations (N up to 1024, K=4, LN=20, Δθ ≥ 9°, M=2000) show RMSE within ~1.5 dB of square Hankel Newton-MUSIC while reducing runtime by more than two orders of magnitude at N=1024 relative to conventional square Hankel MUSIC.

Significance. Single-snapshot high-resolution DoA on large reconfigurable arrays is a genuine bottleneck for real-time ISAC. The combination of fixed-row Hankel truncation with continuous Newton refinement is a practical engineering contribution: the complexity claim is standard matrix arithmetic and is corroborated by the runtime curves in Fig. 2, and the accuracy claim is supported under the stated Monte-Carlo regime. The work supplies a clear algorithmic recipe (Algorithm 1), an explicit complexity table, and reproducible synthetic evaluation against classical baselines. If the accuracy–complexity trade-off extends beyond the comfortable operating point used in the experiments, the method would be immediately useful for large-array sensing; even as a carefully characterized engineering solution it is of interest to the array-signal-processing community.

major comments (2)

The central accuracy claim (RMSE within ~1.5 dB of square Hankel Newton-MUSIC at linear cost) rests on a fixed truncation LN=20 for K=4 and a minimum separation Δθ=9° (Sec. V). Under these settings the noise-subspace dimension is 17 and the sources are well separated relative to the truncated aperture, so the premise that truncation loss is fully recovered by Newton from a 0.5° coarse grid is not stressed. No analytic bound relating admissible K or Δθ to LN is given, and no ablation varies K, LN, or angular separation while keeping N large. When K approaches LN or sources become closer/more unequal, the truncated manifold a_L(θ) loses resolving power and local Newton cannot create peaks that the coarse spectrum never produced. An ablation (or a short theoretical discussion of the effective aperture after truncation) is needed before the claimed trade-off can be regarded as generally esta
Sec. III-B and Algorithm 1 treat K as known and use it both to select the noise subspace dimension and to extract exactly K peaks. In practice K must be estimated from a single snapshot; model-order errors would directly corrupt Un and the subsequent Newton initializations. The manuscript should either include a simple order-selection step (or a sensitivity study) or explicitly state that K is assumed known and discuss the practical implication for ISAC.

minor comments (5)

Figs. 1–2 and the surrounding text repeatedly write “Netown-MUSIC”; correct to “Newton-MUSIC”.
Eq. (7) and the accompanying footnote mix R and RL; a consistent symbol for the smoothed correlation would improve readability.
The abstract and introduction emphasize reconfigurable antennas, yet all experiments use a fixed ULA geometry. A short remark clarifying that the algorithm itself is geometry-agnostic (or a brief non-ULA experiment) would better connect the motivation to the evaluation.
RMSE is reported in dBrad; a brief conversion note or dual axis in degrees would help readers accustomed to degree-scale DoA errors.
A few recent single-snapshot / off-grid DoA references (beyond the cited Hankel-MUSIC and Newtonized OMP works) would better situate the contribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: algorithmic complexity reduction and Monte-Carlo RMSE claims are self-contained and do not reduce to fitted inputs or self-citation chains.

full rationale

This is a standard signal-processing methods paper. The derivation chain is: single-snapshot rank deficiency (Eqs. 3–4) → Hankel spatial smoothing (Eqs. 5–7) → fixed-row truncation LN ≪ N to obtain O(LN²N) correlation/SVD cost (Sec. IV, Table I) → coarse-grid initialization + analytic Newton refinement of the MUSIC cost J(θ) (Eqs. 11–17, Alg. 1). All steps are constructive algorithmic choices or standard calculus on the steering manifold; none is defined in terms of the claimed RMSE or runtime outcome. Empirical claims (RMSE within ~1.5 dB of square Hankel Newton-MUSIC; >100× runtime reduction at N=1024) are obtained from independent synthetic Monte-Carlo trials against classical baselines under stated parameters (N, K=4, LN=20, Δθ=9°, etc.). Self-citations ([6],[7],[15]) supply only ISAC/reconfigurable-antenna context and are not load-bearing for the accuracy or complexity results. There is no fitted physical constant renamed as a prediction, no uniqueness theorem imported from the authors, and no ansatz smuggled via self-citation. Weaknesses (e.g., unstressed regime LN ≫ K, fixed min separation) are correctness/generalization risks, not circularity. Score 0 is therefore the honest finding.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The method rests on the classical far-field ULA single-snapshot model, known source count K, and the empirical choice that a small fixed LN suffices. No new physical entities are postulated. Free parameters are the design knobs LN, Z, and grid densities chosen for the experiments; they are not fitted to hide a theoretical failure but do control the accuracy–complexity trade-off that the central claim depends on.

free parameters (4)

truncation length LN = 20
Hand-chosen LN=20 (with K=4) for all truncated experiments; the linear-complexity claim and the accuracy gap both depend on this choice remaining adequate as N grows.
Newton iteration budget Z = 20
Maximum local iterations set to 20; affects off-grid refinement cost and residual quantization error.
coarse grid spacing ΔΘ_coarse = 0.5°
0.5° coarse grid used to initialize Newton; must be fine enough that each true DoA falls into a basin of attraction of the local quadratic model.
square-Hankel window LO = N/2
Baseline uses LO≈N/2; standard but still a design choice that defines the cubic-cost comparator.

assumptions (6)

domain assumption Far-field narrowband sources observed on a ULA with half-wavelength spacing; steering vector is the standard Vandermonde form (2).
Sec. II-A; entire Hankel construction and manifold derivatives rely on this geometry.
domain assumption Number of sources K is known a priori and LN+1>K so a nonempty noise subspace exists.
Algorithm 1 input and Sec. III-B; MUSIC peak selection and subspace split both require known K.
domain assumption Single temporal snapshot (T=1) with additive white complex Gaussian noise; path gains deterministic or random but fixed per trial.
Sec. II; motivates rank-1 raw covariance and the need for spatial smoothing.
domain assumption Reconfigurable array can be electronically configured into a standard ULA for the sensing slot.
Sec. II opening; abstracts away non-ideal port selection and mutual coupling.
standard math SVD of the smoothed correlation RL yields the same left singular vectors (noise subspace) as SVD of the rectangular Hankel matrix (footnote 1).
Algebraic identity used to justify conventional covariance-style MUSIC on the truncated matrix.
ad hoc to paper Local Newton updates from coarse-grid minima converge to the continuous local minima of J(θ) for the SNRs and separations tested.
Sec. III-B and simulation settings; no basin-of-attraction proof is given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Low-Complexity Gridless Single-Snapshot DoA Estimation via Truncated Hankel Newton-MUSIC." pith.science (2026). https://pith.science/paper/5SY445HQ

@misc{pith2026260708108,
  author       = {Pith},
  title        = {Pith review of: Low-Complexity Gridless Single-Snapshot DoA Estimation via Truncated Hankel Newton-MUSIC},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5SY445HQ}},
  note         = {Machine review of arXiv:2607.08108}
}

read the original abstract

Reconfigurable antenna arrays can provide enhanced spatial Degrees of Freedom (DoFs) for Integrated Sensing And Communication (ISAC) systems, enabling high-resolution Direction of Arrival (DoA) estimation. In highly dynamic scenarios, however, DoA estimation must be performed within short coherence intervals, which often restricts processing to a single snapshot. Conventional subspace methods then suffer from rank deficiency, while Hankel-based spatial smoothing incurs high computational cost when the array size is large. This paper proposes a low-complexity gridless Truncated Hankel NewtonMUSIC framework for single-snapshot DoA estimation. The proposed method constructs a truncated Hankel matrix with a fixed row dimension to recover an effective signal subspace while reducing the cost of correlation construction and subspace decomposition. When the truncation length is independent of the array size, the dominant complexity scales linearly with the number of antenna ports. To reduce grid-induced quantization errors, coarse grid estimates are further refined by a secondorder Newton update in the continuous angular domain. Simulation results show that the proposed method achieves DoA estimation accuracy close to square Hankel Newton-MUSIC while substantially reducing runtime. For large arrays, it provides more than two orders of magnitude runtime reduction compared with conventional square Hankel MUSIC, making it suitable for realtime sensing in reconfigurable antenna-enabled ISAC systems.

Figures

Figures reproduced from arXiv: 2607.08108 by the authors.

**Figure 2.** Average runtime per snapshot versus number of antenn [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages

[1]

Integrated sensing and communications: Toward dual-func tional wire- less networks for 6G and beyond,

F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, an d S. Buzzi, “Integrated sensing and communications: Toward dual-func tional wire- less networks for 6G and beyond,” IEEE J. Sel. Areas Commun. , vol. 40, no. 6, pp. 1728–1767, Mar. 2022

work page 2022
[2]

Collabora tive sensing in perceptive mobile networks: Opportunities and challenges ,

L. Xie, S. Song, Y . C. Eldar, and K. B. Letaief, “Collabora tive sensing in perceptive mobile networks: Opportunities and challenges ,” IEEE Trans. Wireless Commun., vol. 30, no. 1, pp. 16–23, Feb. 2023

work page 2023
[3]

Sensing mutual inform ation with random signals in gaussian channels,

L. Xie, F. Liu, J. Luo, and S. Song, “Sensing mutual inform ation with random signals in gaussian channels,” IEEE Trans. Commun. , vol. 73, no. 10, Oct. 2025

work page 2025
[4]

Fl uid antenna systems,

K.-K. Wong, A. Shojaeifard, K.-F. Tong, and Y . Zhang, “Fl uid antenna systems,” IEEE Trans. Wireless Commun., vol. 20, no. 3, pp. 1950–1962, Mar. 2020

work page 1950
[5]

Movable antennas for wireles s commu- nication: Opportunities and challenges,

L. Zhu, W. Ma, and R. Zhang, “Movable antennas for wireles s commu- nication: Opportunities and challenges,” IEEE Commun. Mag. , vol. 62, no. 6, pp. 114–120, Oct 2023

work page 2023
[6]

Sensing for free: Lear n to localize more sources than antennas without pilots,

W. Y u, K. B. Letaief, and L. Zheng, “Sensing for free: Lear n to localize more sources than antennas without pilots,” IEEE J. Sel. Areas Commun. , vol. 44, pp. 3285–3301, Jan. 2026

work page 2026
[7]

Sensing-Assisted Channel Estimation for Flexible-Antenna Systems: A Unified Framework

R. Cao, W. Y u, Z. Wang, S. Song, J. Zhang, Y . Gong, and K. B. L etaief, “Sensing-assisted channel estimation for ﬂexible-antenn a systems: A uniﬁed framework,” arXiv preprint arXiv:2604.27626 , Apr. 2026

work page Pith review arXiv 2026
[8]

Two decades of array signal proces sing research: the parametric approach,

H. Krim and M. Viberg, “Two decades of array signal proces sing research: the parametric approach,” IEEE Signal Process. Mag. , vol. 13, no. 4, pp. 67–94, Jul. 1996

work page 1996

Show all 15 references

[9]

Multiple emitter location and signal param eter estimation,

R. Schmidt, “Multiple emitter location and signal param eter estimation,” IEEE Trans. Antennas Propag. , vol. 34, no. 3, pp. 276–280, Mar. 1986

1986
[10]

ESPRIT-estimation of signal par ameters via rotational invariance techniques,

R. Roy and T. Kailath, “ESPRIT-estimation of signal par ameters via rotational invariance techniques,” IEEE Trans. Acoust., Speech, Signal Process., vol. 37, no. 7, pp. 984–995, Jul. 1989

1989
[11]

On spatial smoothin g for direction- of-arrival estimation of coherent signals,

T.-J. Shan, M. Wax, and T. Kailath, “On spatial smoothin g for direction- of-arrival estimation of coherent signals,” IEEE Trans. Acoust., Speech, Signal Process. , vol. 33, no. 4, pp. 806–811, Aug. 1985

1985
[12]

MUSIC for single-snapshot sp ectral esti- mation: Stability and super-resolution,

W. Liao and A. Fannjiang, “MUSIC for single-snapshot sp ectral esti- mation: Stability and super-resolution,” Appl. Comput. Harmon. Anal. , vol. 40, no. 1, pp. 33–67, Jan. 2016

2016
[13]

Optimality of Gradien t-MUSIC for spectral estimation,

A. Fannjiang, W. Li, and W. Liao, “Optimality of Gradien t-MUSIC for spectral estimation,” arXiv preprint arXiv:2504.06842 , 2025

2025 arXiv
[14]

Newtonize d orthog- onal matching pursuit: Frequency estimation over the conti nuum,

B. Mamandipoor, D. Ramasamy, and U. Madhow, “Newtonize d orthog- onal matching pursuit: Frequency estimation over the conti nuum,” IEEE Trans. Signal Process. , vol. 64, no. 19, pp. 5066–5081, Oct. 2016

2016
[15]

Joint channel estimation and cooperative lo calization for near-ﬁeld ultra-massive MIMO,

R. Cao, H. He, X. Y u, S. Song, K. Huang, J. Zhang, Y . Gong, a nd K. B. Letaief, “Joint channel estimation and cooperative lo calization for near-ﬁeld ultra-massive MIMO,” arXiv preprint arXiv:2312.13683 , Dec. 2023

2023 arXiv

Pith tools

Reviewed July 10, 2026 · model on record in the stance chip above.

[1] [1]

Integrated sensing and communications: Toward dual-func tional wire- less networks for 6G and beyond,

F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, an d S. Buzzi, “Integrated sensing and communications: Toward dual-func tional wire- less networks for 6G and beyond,” IEEE J. Sel. Areas Commun. , vol. 40, no. 6, pp. 1728–1767, Mar. 2022

work page 2022

[2] [2]

Collabora tive sensing in perceptive mobile networks: Opportunities and challenges ,

L. Xie, S. Song, Y . C. Eldar, and K. B. Letaief, “Collabora tive sensing in perceptive mobile networks: Opportunities and challenges ,” IEEE Trans. Wireless Commun., vol. 30, no. 1, pp. 16–23, Feb. 2023

work page 2023

[3] [3]

Sensing mutual inform ation with random signals in gaussian channels,

L. Xie, F. Liu, J. Luo, and S. Song, “Sensing mutual inform ation with random signals in gaussian channels,” IEEE Trans. Commun. , vol. 73, no. 10, Oct. 2025

work page 2025

[4] [4]

Fl uid antenna systems,

K.-K. Wong, A. Shojaeifard, K.-F. Tong, and Y . Zhang, “Fl uid antenna systems,” IEEE Trans. Wireless Commun., vol. 20, no. 3, pp. 1950–1962, Mar. 2020

work page 1950

[5] [5]

Movable antennas for wireles s commu- nication: Opportunities and challenges,

L. Zhu, W. Ma, and R. Zhang, “Movable antennas for wireles s commu- nication: Opportunities and challenges,” IEEE Commun. Mag. , vol. 62, no. 6, pp. 114–120, Oct 2023

work page 2023

[6] [6]

Sensing for free: Lear n to localize more sources than antennas without pilots,

W. Y u, K. B. Letaief, and L. Zheng, “Sensing for free: Lear n to localize more sources than antennas without pilots,” IEEE J. Sel. Areas Commun. , vol. 44, pp. 3285–3301, Jan. 2026

work page 2026

[7] [7]

Sensing-Assisted Channel Estimation for Flexible-Antenna Systems: A Unified Framework

R. Cao, W. Y u, Z. Wang, S. Song, J. Zhang, Y . Gong, and K. B. L etaief, “Sensing-assisted channel estimation for ﬂexible-antenn a systems: A uniﬁed framework,” arXiv preprint arXiv:2604.27626 , Apr. 2026

work page Pith review arXiv 2026

[8] [8]

Two decades of array signal proces sing research: the parametric approach,

H. Krim and M. Viberg, “Two decades of array signal proces sing research: the parametric approach,” IEEE Signal Process. Mag. , vol. 13, no. 4, pp. 67–94, Jul. 1996

work page 1996