Pith. sign in

REVIEW 3 major objections 5 minor 37 references

Affine Invariant Semi-Blind Receiver: Joint Channel Estimation and High-Order Signal Detection for Multiuser Massive MIMO-OFDM Systems

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read One receiver estimates channel and data for 256-QAM with just two pilots per stream.

desk verdict A promising extension of affine-invariant blind decoding to MU massive MIMO-OFDM, undermined by a data augmentation ablation that cannot work as described. read the letter →

arxiv 2507.21593 v1 pith:SICFQUNJ submitted 2025-07-29 eess.SP

classification eess.SP
keywords semi-blindchannelestimationjointandsignaldetectionmassiveMIMO-OFDMaffineinvarianceconstellationfittinghybridprecoding256-QAMpilotoverheadreduction
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 proposes a semi-blind receiver for downlink multiuser massive MIMO-OFDM that estimates the channel and detects data from the same received symbols, using almost no pilots. It argues that the affine invariance between the transmitted QAM constellation and the received constellation makes this possible: an optimization problem finds a matrix U that maps every received symbol onto the nearest constellation point, and the inverse of U is the channel. The receiver's hybrid precoding suppresses inter-user interference, a few pilots per data stream resolve the inherent permutation and sign ambiguities, and a symmetry-based data augmentation plus iterative refinement keep the method working for 256-QAM over frequency-selective channels. If the result holds, pilot overhead drops from about 67% of resource elements to a few pilot resource elements per user, with an average throughput gain of roughly 11% over standard pilot-based methods.

What carries the argument

The central object is the affine-invariance constellation-fitting program: maximize log|det(U)| subject to $\|UY\|_\infty \leq \lambda_M + 1/\sqrt{\text{SINR}}$, whose optimum is $U = T H_{\text{eq}}^{-1}$, with T a permutation-and-sign ambiguity matrix. The paper's contribution is to make that program work in a massive MIMO-OFDM downlink: a DFT-matrix digital combiner and EZF-based joint transceiver precoder suppress inter-user interference so the residual fits the constraint's SINR margin; a real-valued embedding with block-structure constraint (30) halves the degrees of freedom; orthogonal pilots, one non-zero per stream, initialize the solver and resolve permutation and symbol ambiguity; block-wise processing plus a QAM-symmetry data augmentation ($Y$, $-Y$, $jY$, $-jY$) supply enough samples; and an iterative LS estimation with LLR-based sample rejection followed by LMMSE detection refines the estimates.

What would settle it

Run the method on many independent frequency-selective channel realizations with 256-QAM at high SNR, and compare the SLSQP solution with a globally optimized solution obtained by exhaustive search on a small instance or by many random restarts; if the NMSE or throughput gain over pilot-based methods disappears whenever the true global optimum is used, the reported gain is attributable to initialization luck rather than the constellation-fitting formulation.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that the affine-invariance constellation-fitting idea, previously limited to low-order PAM and block-fading single-user channels, can be extended into a practical semi-blind joint channel estimation and signal detection scheme for downlink MU massive MIMO-OFDM with up to 256-QAM. The transmitter uses a hybrid precoder whose digital part is a joint design with a shared DFT combiner so that the effective channel is nearly diagonalized and residual inter-user interference is small enough to treat as noise. The receiver then solves a nonconvex program that maximizes log|det(U)| under a QAM-adapted infinity-norm constraint; a two-pilot LS estimate provides initialization and identifies which data stream is which, and an algebraic structure constraint halves the optimization degrees of freedom. Frequency selectivity is handled by dividing the time-frequency grid into blocks, and a fourfold data augmentation from QAM's 90-degree rotational symmetry compensates for the smaller sample count. An iterative LS/LMMSE refinement then cleans both channel and symbol estimates. The claimed result is an average throughput gain of about 11% over pilot-based methods, with larger gains at high user density, while supporting 256-QAM.

Load-bearing premise

The load-bearing premise is that the nonconvex optimization, started from the two-pilot LS estimate, reliably reaches the true inverse channel (or a close stationary point) for every block, user, and SNR; the paper only shows the starting point is statistically closer to the optimum than a random one, not that the solver always converges there.

Editorial extensions

If this is right

  • With pilots occupying only $N_s = 2$ resource elements per data stream, pilot overhead no longer grows with user count; the paper reports a theoretical maximum throughput gain of about 20% relative to the 17%-overhead pilot scenario and up to 196% relative to fully orthogonal pilots for 96 streams.
  • The method supports up to 256-QAM on frequency-selective CDL-C channels, whereas the underlying affine-invariance approach was previously limited to low-order PAM.
  • Across SNR levels, the average throughput gain over pilot-based methods is about 11%, and in the 48-user case the gain over non-orthogonal pilots reaches about 20%.
  • Iterative refinement improves channel estimation by roughly 10-15 dB in NMSE at high SNR for lower modulation orders.
  • At high user density, the method is more robust to inter-user interference than non-orthogonal pilot-based estimation, whose NMSE stalls as SNR increases.

Reading between the lines

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

  • A natural extension is to apply the same QAM-symmetry data augmentation to other centrally symmetric constellations such as PSK, where the fourfold rotation may not hold but the reflection structure would.
  • Because the ambiguity-resolving pilots are orthogonal across streams and only one is non-zero, the scheme could be repurposed as a stream-identification layer in any blind or semi-blind equalizer, independent of the constellation-fitting objective.
  • The throughput comparison assumes the nonconvex solver's local convergence; if that assumption weakens at very low SNR or with denser constellations, the practical gain may concentrate in the mid-to-high SNR regime where 256-QAM is actually scheduled.
  • The reported 11% average gain is computed over 20 TTIs with a specific link adaptation table, so a reader should expect the gain to move with the MCS table, the pilot overhead fraction, and the channel model, since the paper's own ablation shows the algorithm's accuracy is modulation-order dependent.
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.

Referee Report

3 major / 5 minor

Summary. The manuscript proposes a semi-blind joint channel estimation and signal detection (JCESD) scheme for downlink multiuser massive MIMO-OFDM. After a hybrid analog/digital precoding design intended to suppress inter-user interference, the paper formulates the estimation problem as a nonconvex constellation-fitting program over an affine transform U, with constraints (24) and (30), initialized by a two-pilot LS estimate. For frequency-selective channels the received matrix is divided into blocks, augmented by rotations of the received samples by -1, +j, -j, and refined by an iterative LS/LMMSE procedure with LLR-based sample removal. Simulations on the Sionna platform compare throughput, NMSE, and BER against orthogonal/non-orthogonal pilot-based CE and EM-based semi-blind CE, reporting an average throughput gain of about 11% and support for 256-QAM.

Significance. If the reported results are correct, the paper has practical value: it demonstrates a way to reduce pilot overhead in heavily loaded MU massive MIMO while supporting high-order modulation. The evaluation includes external baselines, a realistic channel model (CDL-C), and a component-wise ablation, which are strengths. However, the manuscript does not provide code or data, and the central simulation claims rest on the behavior of a nonconvex solver. The most distinctive advertised component, data augmentation, is, as detailed below, inconsistent with the stated optimization problem, and the theoretical pilot-overhead calculation appears to undercount the proposed method's pilots. These issues make the current version unsuitable for acceptance.

major comments (3)
  1. [§III-B1, Fig. 10] Data augmentation as defined by eY_{k,i} = [Y_{k,i}, -Y_{k,i}, jY_{k,i}, -jY_{k,i}] is mathematically inert for the optimization problem (18), (24), (30). For any feasible U, constraint (24) is invariant under multiplication of each column by α ∈ {±1, ±j} because ∥U(αy)∥_∞ = ∥α Uy∥_∞ = ∥Uy∥_∞, and the objective log|det(U)| and the structural constraint (30) are unchanged. Hence the feasible set of the augmented program is identical to that of the original program, and a deterministic solver started from the same U0 returns the same solution. The paper itself concedes that data augmentation does not introduce additional information and 'does not affect the solution of the optimization problem.' It is therefore internally contradictory that Fig. 10 reports that excluding data augmentation yields the most significant NMSE degradation, and that the abstract and contributions list augmentation as a key enabler for high-order modulation. The authors must either identify a mechanism by which augmentation changes the optimization landscape (e.g., a different real-valued representation or a modified constraint) or reconcile the ablation curve with the stated algorithm.
  2. [Appendix A] The convergence analysis in Appendix A bounds the distance between the pilot-based initialization U0 and the global optimum U* under a Neumann-series approximation (Eqs. (42)-(48)), and then cites local convergence of SLSQP (Ref. [37]). This does not prove that SLSQP, from U0, reaches U* or a useful stationary point of the nonconvex program (18), (24), (30). The bound (48) depends on ∥N∥F and requires ∥E∥2 < 1, but no condition is given that guarantees this at the operating SNRs with the small number of pilots used. Since the throughput and NMSE claims of Section IV depend on the solver's global behavior, the paper should either provide a convergence guarantee (or a restart/globalization strategy with a certificate), or report empirical convergence statistics (e.g., the fraction of blocks and users for which the solver reaches a solution consistent with the computed SINR), or state explicitly that the claims are conditioned on successful local convergence.
  3. [§III-D] The pilot-overhead accounting appears inconsistent. The text states that 'Ns pilot symbols are introduced into each data stream' and defines the received pilot block P_{k,r} ∈ C^{Ns×Ns}; with Ns data streams per user, this implies K·Ns^2 pilot REs (e.g., 48·2^2 = 192) for K = 48, Ns = 2. Yet Section III-D and Fig. 4 base the theoretical 196% gain on 'only 2 REs required for pilot transmission regardless of the total number of users.' The comparison should use the actual number of pilot REs consumed by Algorithm 2; otherwise the theoretical maximum throughput gain is overstated. Please clarify what 'Ns=2' counts and re-compute the overhead and the 17% constraint accordingly.
minor comments (5)
  1. [Table I and Section II] The number of OFDM symbols is listed as 'Y = 14' in Table I while the system model uses T; please unify the notation.
  2. [Eq. (21)] Eq. (21) uses \hat X_k before it is defined; specify that it is the hard-decision estimate from the current iteration or from the pilot-based detector.
  3. [Algorithms 1 and 2] The stopping condition ‖U0Yk‖_∞ > 1 in Algorithms 1 and 2 is not consistent with the constraint (24), whose bound is λ_M + sqrt(1/SINR_k); please clarify the normalization assumed in the algorithms.
  4. [Fig. 10 caption] The caption says 'its four simplified variants' but the text lists five removed components (data normalization and initialization, data segmentation, degrees-of-freedom reduction, data augmentation, and original constraints); please make the count consistent.
  5. [Throughout] There are several typographical issues: 'futher' in the Conclusion, 'accurancy' in Section III-D, and 'Reg. Papers1' in Ref. [10]; please proofread the manuscript.

Circularity Check

1 steps flagged · score 4.0 of 10

Central throughput claim is externally benchmarked, but the data-augmentation component is a no-op by the paper's own equations, making its ablation result internally contradictory.

  1. other [Section III-B1 (Data Segmentation and Augmentation); Algorithm 2, step 4; Section IV-C, Fig. 10 ablation discussion]
    "It is important to note that the data augmentation method described above does not introduce additional information. ... This adjustment does not affect the solution of the optimization problem while mitigating the issue of insufficient samples at the boundary of the constellation in high-order modulation. "Notably, excluding the data augmentation mechanism yields the most significant performance loss, underscoring its role in enhancing estimation robustness under sample-scarce conditions.""

    Under (18), (24), and (30), the augmented blocks are rotations of the original block. For any U satisfying the complex-linear structure (30), U(-Y)=-UY, U(jY)=jUY, U(-jY)=-jUY, and the infinity norm in (24) is invariant under sign flips and 90-degree rotations. Every added constraint is therefore termwise identical to a constraint already present, and the objective log|det(U)| is unchanged; the feasible set and the SLSQP solution are exactly the same with or without augmentation. The paper concedes this ('does not affect the solution'), yet Fig. 10 credits the largest NMSE improvement to this mechanism. That attribution reduces by construction to a no-op: the claimed benefit cannot change the optimization outcome, so the ablation result is not reproducible from the stated algorithm.

full rationale

Score 4. The paper's main result, an average ~11% throughput gain over pilot-based and EM-based baselines, comes from simulations against external methods ([22], [34], [36]) and is not circular; it does not reduce to a fitted parameter or to the authors' own prior work. The affine-invariant optimization is adapted from external prior work [20] with new constraints (24)/(30), so the central derivation has independent content. The self-citations present ([3], [27]) are routine background statements and are not load-bearing. However, the data-augmentation mechanism in Section III-B1 is mathematically inert: its added constraints are equivalent to the original ones under the complex-linear structure (30), and the paper explicitly states it 'does not affect the solution.' The Fig. 10 ablation nonetheless attributes the largest NMSE improvement to this mechanism, an attribution that cannot follow from the stated optimization. This is a partial circularity/vacuity in a component-level claim; it does not undermine the external benchmarking of the central throughput claim, but it does make the ablation evidence internally contradictory. No code or data are provided to resolve the contradiction.

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

The method rests on the earlier affine-invariant fitting framework of [20], on an approximate IUI-free precoding model, and on several hand-chosen algorithmic constants. No new physical entities are introduced. The main unverified assumptions are the global convergence of the nonconvex SLSQP optimization, the accuracy of the two-pilot SINR margin, and the claimed benefit of information-free data augmentation.

free parameters (5)
  • data normalization scale a = not specified
    Eq. (20) scales each received stream by a divided by max|y|; the value of a is not given and affects the constraint margin and convergence.
  • LLR decision threshold lambda = 15
    Section III-B2 and Table I set the threshold for discarding low-reliability symbols before LS channel refinement; hand-chosen, not derived.
  • condition number threshold kappa_max = not specified
    Algorithms 1 and 2 return FAIL if kappa(Y_k) > kappa_max; the value is not stated, so the reader cannot reproduce the failure behavior.
  • noise variance floor for LLR = max{10^-3, 1/SNR_k}
    Section III-B2 introduces a heuristic lower bound to avoid numerical overflow; this is a hand-chosen constant affecting symbol reliability filtering.
  • number of blocks Nf and iterations Niter = Nf=8, Niter=5
    Table I sets these; they are design choices that trade complexity against accuracy, not derived from a principle.
assumptions (5)
  • domain assumption Hybrid precoding with a random DFT combiner W_BB and EZF precoding from the mean Gram matrix (8)-(12) approximately diagonalizes the multiuser channel, leaving residual IUI that can be treated as zero-mean Gaussian noise.
    Section II-A2 assumes the joint transceiver precoding makes \bar{V}_k F_{BB,k} approximately I and \bar{V}_k F_{BB,i} approximately O; this is an approximation over frequency-selective channels and is not proven.
  • domain assumption The l_infinity constraint margin sqrt(1/SINR_k) estimated from Ns=2 pilot symbols in (22)-(23) correctly bounds residual IUI plus AWGN for all samples in the block.
    Section III-A2 replaces the fixed noise margin c*sigma in [20] with an SINR-based margin; the SINR estimate from two pilots is assumed accurate enough.
  • ad hoc to paper SLSQP, started from the pilot-based initialization, converges to the global optimum of the nonconvex program (18), (24), (30), or to a stationary point close enough to yield accurate CE/SD.
    Appendix A analyzes only the distance between initial and optimal U; it cites [37] for local convergence but does not establish the basin of attraction or global convergence.
  • ad hoc to paper Data augmentation by rotating received samples by 0, 90, 180, 270 degrees increases the effective number of samples without changing the solution of the optimization program.
    Section III-B1 states augmentation 'does not introduce additional information' but relies on it to mitigate sample scarcity at constellation boundaries; the mechanism is not formalized.
  • standard math QAM constellation symmetry: for every constellation point x, -x, jx, -jx are also constellation points.
    Used in Eq. (31)-(32) for data augmentation; standard property of square QAM constellations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Affine Invariant Semi-Blind Receiver: Joint Channel Estimation and High-Order Signal Detection for Multiuser Massive MIMO-OFDM Systems." pith.science (2026). https://pith.science/paper/SICFQUNJ

@misc{pith2026250721593,
  author       = {Pith},
  title        = {Pith review of: Affine Invariant Semi-Blind Receiver: Joint Channel Estimation and High-Order Signal Detection for Multiuser Massive MIMO-OFDM Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SICFQUNJ}},
  note         = {Machine review of arXiv:2507.21593}
}
read the original abstract

Massive multiple input and multiple output (MIMO) systems with orthogonal frequency division multiplexing (OFDM) are foundational for downlink multi-user (MU) communication in future wireless networks, for their ability to enhance spectral efficiency and support a large number of users simultaneously. However, high user density intensifies severe inter-user interference (IUI) and pilot overhead. Consequently, existing blind and semi-blind channel estimation (CE) and signal detection (SD) algorithms suffer performance degradation and increased complexity, especially when further challenged by frequency-selective channels and high-order modulation demands. To this end, this paper proposes a novel semi-blind joint channel estimation and signal detection (JCESD) method. Specifically, the proposed approach employs a hybrid precoding architecture to suppress IUI. Furthermore we formulate JCESD as a non-convex constellation fitting optimization exploiting constellation affine invariance. Few pilots are used to achieve coarse estimation for initialization and ambiguity resolution. For high-order modulations, a data augmentation mechanism utilizes the symmetry of quadrature amplitude modulation (QAM) constellations to increase the effective number of samples. To address frequency-selective channels, CE accuracy is then enhanced via an iterative refinement strategy that leverages improved SD results. Simulation results demonstrate an average throughput gain of 11\% over widely used pilot-based methods in MU scenarios, highlighting the proposed method's potential to improve spectral efficiency.

Figures

Figures reproduced from arXiv: 2507.21593 by the authors.

Figure 1
Figure 1. Structure of the hybrid precoder and combiner at the transmitter and receivers, where the RF precoder and combiner adopt a sub-connected phase [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The method of eliminating ambiguity using few pilots. Orthogonal pilots are employed between different data streams of each user. Each data stream [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. The division of the received symbols from the whole TTI into multiple [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The pilot pattern specified by the 3GPP standard with 96 streams [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: The throughput of the proposed algorithm: (a) 24 users transmitting 48 streams, (b) 48 users transmitting 96 streams. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: The NMSE of the proposed algorithm: (a) 24 users transmitting 48 streams, (b) 48 users transmitting 96 streams. [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: The BER of the proposed algorithm: (a) 24 users transmitting 48 streams, (b) 48 users transmitting 96 streams. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: The NMSE of the proposed algorithm under different modulation [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 10
Figure 10. Figure 10: NMSE performance comparison of the proposed method without [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 36 canonical work pages

  1. [37]

    Nocedal and S

    J. Nocedal and S. J. Wright, Numerical Optimization , New York, NY , USA: Springer, 2006

  2. [1]

    Estimation of broadband multiuser millimeter wave massive MIMO-OFDM channels by exploiting their sparse structure,

    X. Lin, S. Wu, C. Jiang, L. Kuang, J. Yan and L. Hanzo, “Estimation of broadband multiuser millimeter wave massive MIMO-OFDM channels by exploiting their sparse structure,” IEEE Trans. Wireless Commun. , vol. 17, no. 6, pp. 3959-3973, Jun. 2018

  3. [2]

    Cell-free massive MIMO-OFDM trans- mission over frequency-selective fading channels,

    W. Jiang and H. D. Schotten, “Cell-free massive MIMO-OFDM trans- mission over frequency-selective fading channels,” IEEE Commun. Lett., vol. 25, no. 8, pp. 2718-2722, Aug. 2021

  4. [3]

    Low-complexity iterative detection for large-scale multiuser MIMO-OFDM systems using approximate message passing,

    S. Wu, L. Kuang, Z. Ni, J. Lu, D. Huang and Q. Guo, “Low-complexity iterative detection for large-scale multiuser MIMO-OFDM systems using approximate message passing,” IEEE J. Sel. Topics Signal Process , vol. 8, no. 5, pp. 902-915, Oct. 2014

  5. [4]

    Inter-user interference suppression in time varying channel with null- space expansion for multiuser massive MIMO,

    T. Iwakuni, K. Maruta, A. Ohta, Y . Shirato, T. Arai and M. Iizuka, “Inter-user interference suppression in time varying channel with null- space expansion for multiuser massive MIMO,” 2015 IEEE 26th Annual International Symposium on Personal, Indoor, and Mobile Radio Com- munications (PIMRC), Hong Kong, China, 2015, pp. 558-562

  6. [5]

    Energy and spectral effi- ciency of very large multiuser MIMO systems,

    H. Q. Ngo, E. G. Larsson and T. L. Marzetta, “Energy and spectral effi- ciency of very large multiuser MIMO systems,” IEEE Trans. Commun., vol. 61, no. 4, pp. 1436-1449, Apr. 2013

  7. [6]

    Downlink user capacity of massive MIMO under pilot contamination,

    J. -C. Shen, J. Zhang and K. B. Letaief, “Downlink user capacity of massive MIMO under pilot contamination,” IEEE Trans. Wireless Commun., vol. 14, no. 6, pp. 3183-3193, Jun. 2015

  8. [7]

    NR; physical channels and modulation (release 18),

    3GPP TS 38.211, “NR; physical channels and modulation (release 18),” 3rd Generation Partnership Project; Technical Specification Group Radio Access Network

Show all 37 references
  1. [8]

    Blind channel estimation and data detection with unknown modulation and coding scheme,

    Y . Liu and F. Wang, “Blind channel estimation and data detection with unknown modulation and coding scheme,” IEEE Trans. Commun. , vol. 72, no. 5, pp. 2595-2609, May 2024

  2. [9]

    Efficient blind channel estimation for IRS-assisted joint communication and sensing in 6G networks,

    A. A. Siddig, et al., “Efficient blind channel estimation for IRS-assisted joint communication and sensing in 6G networks,” 2024 IEEE Middle East Conference on Communications and Networking (MECOM) , Abu Dhabi, United Arab Emirates, 2024, pp. 309-314

  3. [10]

    Indeterminacy and identifiability of blind identification,

    L. Tong, R. . -w. Liu, V . C. Soon and Y . . -F. Huang, “Indeterminacy and identifiability of blind identification,” IEEE Trans. Circuits Syst. I, Reg. Papers1, vol. 38, no. 5, pp. 499-509, May 1991

  4. [11]

    Fast semi-blind channel estimation for MIMO- OFDM systems with virtual carriers,

    J.-L. Yu and W.-R. Kuo, “Fast semi-blind channel estimation for MIMO- OFDM systems with virtual carriers,” 2012 1st IEEE International Conference on Communications in China (ICCC) , Beijing China, 2012, pp. 356-361

  5. [12]

    Fast subspace-based semi-blind channel estimation for MIMO-OFDM com- munications,

    K. N. Aliyu, A. Lawal, K. Abed-Meraim and A. Zerguine, “Fast subspace-based semi-blind channel estimation for MIMO-OFDM com- munications,” 2023 31st European Signal Processing Conference (EU- SIPCO), Helsinki, Finland, 2023, pp. 1460-1463

  6. [13]

    Semi-blind channel estimation for DCO-OFDM VLC systems,

    Y . Li, Y . Jiang, X. Zhu, S. Sun and V . K. N. Lau, “Semi-blind channel estimation for DCO-OFDM VLC systems,” ICC 2024 - IEEE International Conference on Communications, Denver, CO, USA, 2024, pp. 3707-3712

  7. [14]

    Second- order statistics-based semi-blind techniques for channel estimation in millimeter-wave MIMO analog and hybrid beamforming,

    P. Singh, S. Srivastava, A. K. Jagannatham and L. Hanzo, “Second- order statistics-based semi-blind techniques for channel estimation in millimeter-wave MIMO analog and hybrid beamforming,” IEEE Trans. Commun., vol. 68, no. 11, pp. 6886-6901, Nov. 2020

  8. [15]

    Blind channel estimation for CP/CP-free OFDM systems using subspace approach,

    S. -H. Fang, J. -Y . Chen, J. -S. Lin, M. -D. Shieh and J. -Y . Hsu, “Blind channel estimation for CP/CP-free OFDM systems using subspace approach,” 2015 IEEE 81st Vehicular Technology Conference (VTC Spring), Glasgow, UK, 2015, pp. 1-5

  9. [16]

    A per- formance analysis of optimized semi-blind channel estimation method in OFDM systems,

    S. Gulomjon, F. Yongqing, J. Sangirov, F. Ye and A. Olmasov, “A per- formance analysis of optimized semi-blind channel estimation method in OFDM systems,” 2017 19th International Conference on Advanced Communication Technology (ICACT) , PyeongChang, Korea (South), 2017, pp. 907-912

  10. [17]

    Reliable blind channel estimation scheme based on cross-correlated cyclic prefix for OFDM system,

    Taejoon Kim and Iksoo Eo, “Reliable blind channel estimation scheme based on cross-correlated cyclic prefix for OFDM system,” 2006 8th In- ternational Conference Advanced Communication Technology , Phoenix Park, Korea (South), 2006, pp. 3-5

  11. [18]

    Fast subspace-based blind and semi-blind channel estimation for MIMO-OFDM Systems,

    O. Rekik, K. N. Aliyu, B. M. Tuan, K. Abed-Meraim and N. L. Trung, “Fast subspace-based blind and semi-blind channel estimation for MIMO-OFDM Systems,” IEEE Trans. Wireless Commun. , vol. 23, no. 8, pp. 10247-10257, Aug. 2024

  12. [19]

    Fast convergence on blind and semi-blind channel estimation for MIMO–OFDM systems,

    Zhang B, Yu J L and Kuo W, “Fast convergence on blind and semi-blind channel estimation for MIMO–OFDM systems,” Circuits, Systems, and Signal Processing, vol. 34, pp. 1993-2013, Jun. 2015

  13. [20]

    Blind joint MIMO channel estimation and decoding,

    T. R. Dean, M. Wootters and A. J. Goldsmith, “Blind joint MIMO channel estimation and decoding,” IEEE Trans. Inf. Theory , vol. 65, no. 4, pp. 2507-2524, Apr. 2019

  14. [21]

    Fast blind MIMO decoding through vertex hopping,

    T. R. Dean, J. R. Perlstein, M. Wootters and A. J. Goldsmith, “Fast blind MIMO decoding through vertex hopping,” IEEE Trans. Wireless Commun., vol. 18, no. 7, pp. 3669-3682, Jul. 2019

  15. [22]

    Semi-blind channel estimation for multiuser massive MIMO systems,

    E. Nayebi and B. D. Rao, “Semi-blind channel estimation for multiuser massive MIMO systems,” IEEE Trans. Signal Process. , vol. 66, no. 2, pp. 540-553, Jan. 2018

  16. [23]

    An overview of signal processing techniques for millimeter wave MIMO systems,

    R. W. Heath, N. Gonz ´alez-Prelcic, S. Rangan, W. Roh and A. M. Sayeed, “An overview of signal processing techniques for millimeter wave MIMO systems,” IEEE J. Sel. Topics Signal Process. , vol. 10, no. 3, pp. 436–453, Apr. 2016

  17. [24]

    A dynamic array-of-subarrays architecture and hybrid precoding algorithms for terahertz wireless communications,

    L. Yan, C. Han, and J. Yuan, “A dynamic array-of-subarrays architecture and hybrid precoding algorithms for terahertz wireless communications,” IEEE J. Sel. Areas Commun. , vol. 38, no. 9, pp. 2041–2056, Sep. 2020

  18. [25]

    Hybrid digital and analog beamforming design for large-scale antenna arrays,

    F. Sohrabi and W. Yu, “Hybrid digital and analog beamforming design for large-scale antenna arrays,” IEEE J. Sel. Topics Signal Process., vol. 10, no. 3, pp. 501-513, Apr. 2016

  19. [26]

    Eigen-based transceivers for the MIMO broadcast channel with semi-orthogonal user selection,

    L. Sun and M. R. McKay, “Eigen-based transceivers for the MIMO broadcast channel with semi-orthogonal user selection,” IEEE Trans. Signal Process., vol. 58, no. 10, pp. 5246-5261, Oct. 2010

  20. [27]

    Dual-driven learning for channel estimation of massive MIMO systems with one-bit ADCs,

    H. Li, S. Wu, S. Zheng, C. Jiang, H. Yang and W. Zhang, “Dual-driven learning for channel estimation of massive MIMO systems with one-bit ADCs,” IEEE Wireless Commun. Lett. , vol. 13, no. 4, pp. 1108-1112, Apr. 2024

  21. [28]

    Performance analysis and comparisons of hybrid precoding scheme for multi-user mmwave massive MIMO system,

    T. A. Abose, T. O. Olwal, M. R. Hassen and E. S. Bekele, “Performance analysis and comparisons of hybrid precoding scheme for multi-user mmwave massive MIMO system,” 2022 3rd International Conference for Emerging Technology (INCET) , Belgaum, India, 2022, pp. 1-6

  22. [29]

    Frequency-domain compressive channel estimation for frequency-selective hybrid millimeter wave MIMO systems,

    J. Rodr ´ıguez-Fern´andez, N. Gonz ´alez-Prelcic, K. Venugopal and R. W. Heath, “Frequency-domain compressive channel estimation for frequency-selective hybrid millimeter wave MIMO systems,” IEEE Trans. Wireless Commun., vol. 17, no. 5, pp. 2946-2960, May 2018

  23. [30]

    Analysis of different planar antenna arrays for mmwave massive MIMO systems,

    W. Tan, S. D. Assimonis, M. Matthaiou, Y . Han, X. Li and S. Jin, “Analysis of different planar antenna arrays for mmwave massive MIMO systems,” 2017 IEEE 85th Vehicular Technology Conference (VTC Spring), Sydney, NSW, Australia, 2017, pp. 1-5

  24. [31]

    Blind separation of synchronous co-channel digital signals using an antenna array. I. Algorithms,

    S. Talwar, M. Viberg and A. Paulraj, “Blind separation of synchronous co-channel digital signals using an antenna array. I. Algorithms,” IEEE Trans. Signal Process., vol. 44, no. 5, pp. 1184-1197, May 1996

  25. [32]

    A software package for sequential quadratic programming,

    D. Kraft, “A software package for sequential quadratic programming,” Deutsche Forschungs- und Versuchsanstalt f ¨ur Luft- und Raumfahrt , Report DFVLR-FR 88-28, 1988

  26. [33]

    pyOpt: A Python- based object-oriented framework for nonlinear constrained optimiza- tion,

    R. E. Perez, P. W. Jansen and J. R. R. A. Martins, “pyOpt: A Python- based object-oriented framework for nonlinear constrained optimiza- tion,” Structures Multidiscipl. Optim., vol. 45, no. 1, pp. 101–118, 2012

  27. [34]

    NR; physical layer procedures for data (release 18),

    3GPP TS 38.214, “NR; physical layer procedures for data (release 18),” 3rd Generation Partnership Project; Technical Specification Group Radio Access Network

  28. [35]

    Sionna: an open-source library for next-generation physical layer research

    J. Hoydis et al. “Sionna: an open-source library for next-generation physical layer research”, 2022, arXiv: 2203.11854

  29. [36]

    On channel estimation in OFDM systems

    J. Beek, O. Edfors, M. Sandell, S. K. Wilson and P. o. Borjesson, “On channel estimation in OFDM systems”, Proc. of IEEE 45th Vehicular Technology Conference, pp. 815-819, July 1995

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.