Pith. sign in

REVIEW 3 major objections 38 references

Multi-base multi-scale time-frequency fusion lets an ISAC network tell drones from birds, cars and people with near-perfect accuracy.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-14 12:39 UTC pith:4NWXVQKW

load-bearing objection Solid multi-BS multi-scale TF fusion recipe plus a 1.44 M synthetic LAT benchmark; accuracies are real inside free-space simulation but remain optimistic upper bounds until real channels appear. the 3 major comments →

arxiv 2607.10319 v1 pith:4NWXVQKW submitted 2026-07-11 eess.SP

Networked ISAC Enabled Target Recognition Towards Low-Altitude Economy

Hongliang Luo , Chuanbin Zhao , Boxuan Sun , Zhonghua Chu , Shengjie Quan , Guangyi Liu , Feifei Gao This is my paper
classification eess.SP
keywords integrated sensing and communicationslow-altitude economylow-altitude target recognitionmulti-scale time-frequency featuremulti-base-station collaborationSwin Transformermicro-Doppler
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Low-altitude monitoring with 6G-style integrated sensing and communications stations is plagued by false alarms: birds in the main beam and vehicles or pedestrians in the vertical sidelobes produce echoes that look like drones. This paper claims the cure is to treat recognition as a multi-view, multi-resolution problem. From each base station’s echo it extracts three complementary time-frequency images—one tuned for velocity detail, one for temporal micro-motion rhythm, and one that tracks velocity transitions via matched filtering—then fuses those images across three cooperating stations inside a hierarchical Swin Transformer. On a new 1.44-million-sample simulated benchmark the network reaches 99.88 % average accuracy when subtypes are seen in training and still 97.82 % when entirely new subtypes appear, outperforming any single feature or single station. The result matters because it shows that scarce millisecond-scale sensing resources and modest arrays can still support reliable low-altitude security once complementary observations are deliberately engineered and fused.

Core claim

The authors establish that fusing three deliberately complementary time-frequency spectra (velocity-resolution-preferred, time-resolution-preferred, and velocity-transfer) across multiple cooperating base stations yields high recognition accuracy and strong unseen-subtype generalization for the four-class problem of UAV versus bird versus vehicle versus pedestrian under practical sub-6 GHz ISAC resource constraints.

What carries the argument

Multi-scale feature of the LAT: the triple of VRP-TF, TRP-TF and VT-TF spectra extracted per base station, visualized as images, then processed by shared Swin-B backbones followed by intra-BS and inter-BS transformer fusion.

Load-bearing premise

The entire claim rests on idealized free-space point-cloud scattering models and perfect range-bin selection; real multipath, hardware impairments and clutter are never present in the 1.44 million training or test samples.

What would settle it

Collect real sub-6 GHz multi-BS ISAC echoes of drones, birds, vehicles and pedestrians under the same sparse-symbol regime, retrain or evaluate the identical multi-scale fusion network, and check whether the reported accuracy and unseen-subtype gap survive once multipath and clutter appear.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • ISAC base stations can suppress the dominant non-UAV false alarms that currently limit low-altitude security without needing denser time-frequency resources.
  • The 1.44-million-sample multi-BS multi-scale dataset becomes a public benchmark for subsequent LAT recognition algorithms.
  • Hierarchical intra- then inter-station fusion is a reusable template for other multi-view sensing tasks that face resolution trade-offs.
  • Unseen-subtype generalization above 97 % average accuracy suggests the features capture category-level micro-motion physics rather than subtype idiosyncrasies.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the same multi-scale spectra remain discriminative under measured multipath, the method could be retrofitted onto existing 5G NR sparse sensing patterns with only software changes.
  • The velocity-transfer spectrum construction may transfer to other micro-Doppler problems (e.g., gesture or gait recognition) where classical STFT forces an awkward resolution trade-off.
  • Extending the four-class taxonomy to include fixed-wing UAVs or multi-rotor swarms would test whether the fusion still separates overlapping micro-motion signatures.

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

3 major / 0 minor

Summary. The paper proposes a multi-base-station (multi-BS), multi-scale feature-fusion scheme for low-altitude target (LAT) recognition in sub-6 GHz ISAC networks. It formulates multi-scatterer motion and free-space echo models for UAV, bird, vehicle, and pedestrian under a three-BS cell; extracts three time-frequency representations per BS (velocity-resolution-preferred STFT, time-resolution-preferred STFT, and a matched-filter velocity-transfer spectrum); and fuses the corresponding RGB visualizations with a shared Swin-B backbone plus intra-BS and inter-BS Transformer encoders. A synthetic dataset of 1.44 million multi-BS samples is released as a public benchmark. Simulation ablations report 99.88% average accuracy in the seen-subtype setting and 97.82% in the unseen-subtype setting, with consistent gains from multi-scale fusion and multi-BS collaboration over single-feature and single-BS baselines, and modest gains over mean-fusion Swin-B/ConvNeXt/ViT variants.

Significance. If the simulation claims hold under the stated free-space model, the work is a solid and timely contribution to ISAC-enabled low-altitude monitoring. Distinguishing UAVs from birds, vehicles, and pedestrians is a practically important false-alarm problem that prior ISAC recognition work largely under-addresses. The multi-scale feature set (especially the VT-TF construction), the hierarchical multi-BS fusion architecture, the large multi-class multi-BS dataset with an explicit unseen-subtype protocol, and the public project page are concrete strengths that can serve as a reusable benchmark for the community. The ablations in Tables II–IV cleanly isolate multi-scale and multi-BS gains. The main caveat is that all evidence is synthetic under idealized channels; the paper’s value is therefore primarily as a controlled methodological and dataset contribution rather than as a validated field system.

major comments (3)
  1. Section III-E (Eq. 24 and the following footnote) and Section IV-A assume free-space multi-scatterer channels, perfect erase of the known sensing symbol, and perfect range-bin selection via argmax after IDFT. Multipath, clutter, hardware impairments, and imperfect bin selection are deferred to future work and never injected into the 1.44 M training or test samples. The reported 99.88%/97.82% figures are therefore upper bounds inside an idealized world. For a paper framed toward low-altitude security and false-alarm mitigation, this is load-bearing for external validity. Please either (i) add at least one controlled robustness study (e.g., residual multipath, range-bin offset, or unequal multi-BS SNR) showing how VRP/TRP/VT features and multi-BS fusion degrade, or (ii) substantially expand the limitations discussion with a concrete analysis of which of the three spectra are expected to be
  2. Tables II–IV compare feature combinations, single- vs multi-BS, and backbone/fusion variants, but do not compare against prior LAT recognition pipelines (e.g., micro-Doppler/STFT classifiers or networks from [21]–[24], even in single-BS form on the new four-class data). Without such baselines it is hard to separate the contribution of the proposed multi-scale features and multi-BS fusion from the general power of a strong ImageNet-pretrained vision backbone on clean synthetic spectrograms. Please add at least one literature-style baseline (or a carefully justified statement why adaptation is infeasible) so that the incremental value of VRP+TRP+VT and of inter-BS fusion is measured against existing LAT recognition practice, not only against mean fusion of the same features.
  3. Section IV-E, Eqs. (30)–(35): the VT-TF “velocity transition amount” G_{k,g,i} is defined as a squared difference of matched-filter magnitudes on adjacent windows. The text asserts that this captures periodic velocity/acceleration transfer and is complementary to STFT, but provides little analysis of what physical quantity is being estimated, its bias under multi-scatterer superposition, or why the particular normalization and squaring are preferred. Because VT-TF is presented as a core novel multi-scale feature, please add a short derivation or interpretive argument (and, if possible, a simple single-scatterer sanity check) clarifying what G_{k,g,i} measures and when it is expected to outperform ordinary short-window STFT.

Circularity Check

0 steps flagged

No significant circularity: accuracies are empirical results on held-out simulated samples with fixed (non-fitted) multi-scale extractors; self-citations supply prior monostatic models but do not force the multi-BS fusion numbers.

full rationale

The paper's load-bearing claims are the simulated recognition accuracies (Tables II–IV: 99.88 % seen-subtype avg, 97.82 % unseen-subtype avg) obtained by training a Swin-B fusion network on visualized VRP-TF / TRP-TF / VT-TF images and evaluating on held-out samples (including completely unseen subtypes). Feature extraction uses fixed STFT window/FFT parameters and a fixed matched-filter velocity-transfer construction (Eqs. 28–35); none of these parameters are fitted to the class labels. Motion/echo models (Eqs. 7, 14–24) are physics-based point-cloud constructions (Rodrigues rotation, attitude matrices, CMU mocap interpolation) that generate the synthetic dataset; they do not algebraically encode the final accuracy figures. Self-citation of the authors’ prior monostatic work [24] is used only to reuse UAV/bird motion equations and to motivate the problem; the multi-BS geometry, vehicle/pedestrian models, three-scale features, hierarchical intra-/inter-BS transformers, and the 1.44 M-sample multi-BS dataset are new and independently ablated. No equation reduces a claimed accuracy or “prediction” to an input by construction, no uniqueness theorem is imported, and no free parameter is fitted then re-presented as a forecast. The sole idealization (free-space channels, perfect range-bin selection) is explicitly flagged and does not create internal circularity. Hence the derivation chain is self-contained against its own synthetic benchmarks.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 3 invented entities

The central performance claims rest on a large set of hand-chosen simulation parameters and standard radar/OFDM modeling assumptions; the only truly new constructs are the three named multi-scale spectra and the fusion network itself. No free parameters are fitted to real measurements because no real measurements are used.

free parameters (5)
  • STFT window lengths and FFT sizes (L_win,long=256, L_win,short=8, P_win,short=1024, etc.)
    Chosen by hand to realize the velocity- vs time-resolution trade-off; directly determine the three input images.
  • VT-TF velocity grid (V_win,min=-6.1, V_win,max=6.1, ΔV=0.1)
    Hand-selected search range and step that define the third feature map.
  • TRF parameters (Tr=2.5 ms, N=400 symbols, M=256 subcarriers, array 16×4)
    Resource-allocation choices that set the sampling of every echo and therefore every spectrum.
  • SNR set {3,8,13,18,23,∞} dB and equal-SNR multi-BS assumption
    Noise levels injected into the synthetic dataset; performance tables are conditioned on them.
  • Subtype split (5 seen / 5 unseen per class) and 6000 motion models per subtype
    Determines the generalization numbers reported in Tables III–IV.
axioms (5)
  • domain assumption LAT echo channel is the coherent sum of L independent scattering-point channels (Eq. 7) whose kinematics fully encode micro-Doppler.
    Standard point-cloud radar abstraction; used for all four target classes.
  • domain assumption OFDM symbol-domain complex signal after perfect range-bin selection (IDFT + argmax) is a sufficient statistic for recognition.
    Section IV-A; ignores residual multipath and bin leakage.
  • domain assumption Rodrigues rotation, attitude matrices and CMU mocap interpolation correctly generate realistic micro-motions for UAV paddles, bird wings, vehicle wheels and pedestrian limbs.
    Sections III-A–D; foundation of the entire synthetic dataset.
  • ad hoc to paper ImageNet-pretrained Swin-B backbone transfers useful features to visualized TF spectra after 224×224 resize.
    Architectural choice in Section IV-F; not derived from first principles.
  • standard math Standard STFT and matched-filter mathematics (Eqs. 25–35).
    Used without modification to produce the three spectra.
invented entities (3)
  • Velocity-transfer time-frequency (VT-TF) spectrum no independent evidence
    purpose: Third multi-scale feature that captures periodic velocity-component transitions via matched-filter differencing.
    Defined in Section IV-E; no prior independent measurement of this exact construct is cited.
  • Multi-BS multi-scale feature fusion LAT recognition network (Swin-B + intra/inter-station transformers) no independent evidence
    purpose: Joint classifier that fuses nine TF images into a four-class decision.
    Architecture of Section IV-F and Fig. 9; performance is the paper’s main empirical claim.
  • 1.44-million-sample multi-BS LAT echo dataset no independent evidence
    purpose: Public benchmark for training and evaluating LAT recognition algorithms.
    Generated entirely by the authors’ simulators; claimed as community resource.

pith-pipeline@v1.1.0-grok45 · 28025 in / 3438 out tokens · 36173 ms · 2026-07-14T12:39:19.188886+00:00 · methodology

0 comments
read the original abstract

In this paper, we propose a low-altitude target (LAT) recognition scheme based on multi-base station (BS) collaboration and multi-scale feature fusion for integrated sensing and communications (ISAC) network. Firstly, we formulate the motion equations, echo channels, and echo signals for unmanned aerial vehicle (UAV), bird, vehicle, and pedestrian under multi-BS collaborative monitoring scenario. Then we extract the velocityresolution-preferred time-frequency spectrum, time-resolutionpreferred time-frequency spectrum, and the velocity-transfer time-frequency spectrum observed by each BS from echo signals. We collectively refer to these three types of time-frequency spectrum as the multi-scale feature of the LAT. Next, we design a multi-BS and multi-scale feature fusion enabled LAT recognition network with Swin Transformer, which employs the visualized images of multi-scale feature to jointly recognize the target through deep feature extraction, intra-BS feature interaction, inter-BS feature interaction, and target recognition output. We generate a massive echo signal dataset comprising 1,440,000 samples for LAT recognition within ISAC network. This dataset can serve as a public benchmark to evaluate our proposed scheme and facilitate future research. Simulation results demonstrate that the proposed scheme realizes high recognition accuracy and robust unseen-subtype generalization, confirming the effectiveness of multi-scale feature fusion and the additional gains brought by multi-BS collaboration. The project page is available at: https://alivn999.github.io/COSMOS-Networked-ISAC-Enabl ed-Target-Recognition-Towards-Low-Altitude-Economy/.

Figures

Figures reproduced from arXiv: 2607.10319 by Boxuan Sun, Chuanbin Zhao, Feifei Gao, Guangyi Liu, Hongliang Luo, Shengjie Quan, Zhonghua Chu.

Figure 1
Figure 1. Figure 1: LAT recognition scenario in ISAC network. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Aerial view of multi-BS cooperative sensing scenario. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Process diagram to obtain the vehicle’s calibration point cloud. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Examples of pedestrian point cloud in CMU dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: VRP-TF spectrum observed by BS-A for four categories of LATs, in which the horizontal axis represents the time, and the vertical axis represents the Doppler units. (a) UAV. (b) Bird. (c) Vehicle. (d) Pedestrian. IV. MULTI-BS AND MULTI-SCALE FEATURE FUSION ENABLED LOW-ALTITUDE TARGET RECOGNITION SCHEME In this section, we propose the multi-BS and multi-scale feature fusion enabled LAT recognition scheme. A.… view at source ↗
Figure 7
Figure 7. Figure 7: TRP-TF spectrum observed by BS-A for four categories of LATs, in which the horizontal axis represents the time, and the vertical axis represents [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: VT-TF spectrum observed by BS-A for four categories of LATs, in which the horizontal axis represents the time, and the vertical axis represents the [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Multi-BS and multi-scale feature fusion enabled LAT recognition network based on Swin-B. [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 2 linked inside Pith

  1. [1]

    The potential of low-altitude airspace: The future of urban air transportation,

    H. Huang, J. Su, and F.-Y . Wang, “The potential of low-altitude airspace: The future of urban air transportation,”IEEE Trans. Intell. V eh., vol. 9, no. 8, pp. 5250–5254, Aug. 2024

    2024

  2. [2]

    Random signal design for joint communication and SAR imaging towards low-altitude economy,

    B. Zheng and F. Liu, “Random signal design for joint communication and SAR imaging towards low-altitude economy,”IEEE Wireless Com- mun. Lett., vol. 13, no. 10, pp. 2662–2666, Oct. 2024

    2024

  3. [3]

    Low-altitude intelligent transportation: System architecture, infrastructure, and key technologies,

    C. Huang, S. Fang, H. Wu, Y . Wang, and Y . Yang, “Low-altitude intelligent transportation: System architecture, infrastructure, and key technologies,”J. Ind. Inf. Integr ., vol. 42, p. 100694, Nov. 2024

    2024

  4. [4]

    6G non-terrestrial networks enabled low- altitude economy: Opportunities and challenges,

    Y . Jianget al., “6G non-terrestrial networks enabled low- altitude economy: Opportunities and challenges,”arXiv e-prints, p. arXiv:2311.09047, Nov. 2023

    Pith/arXiv arXiv 2023

  5. [5]

    Autonomous eVTOL: A summary of researches and challenges,

    S. Xiang, A. Xie, M. Ye, X. Yan, X. Han, H. Niu, Q. Li, and H. Huang, “Autonomous eVTOL: A summary of researches and challenges,”Green Energy Intell. Transp., vol. 3, no. 1, p. 100140, Feb. 2024

    2024

  6. [6]

    Survey on collaborative smart drones and internet of things for improving smartness of smart cities,

    S. H. Alsamhi, O. Ma, M. S. Ansari, and F. A. Almalki, “Survey on collaborative smart drones and internet of things for improving smartness of smart cities,”IEEE Access, vol. 7, pp. 128125–128152, Aug. 2019

    2019

  7. [7]

    IoTaaS: drone-based internet of things as a service framework for smart cities,

    M. A. Hoqueet al., “IoTaaS: drone-based internet of things as a service framework for smart cities,”IEEE Internet Things J., vol. 9, no. 14, pp. 12425–12439, Jul. 2022

    2022

  8. [8]

    Collaboration of drone and internet of public safety things in smart cities: An overview of QoS and network performance optimization,

    S. H. Alsamhiet al., “Collaboration of drone and internet of public safety things in smart cities: An overview of QoS and network performance optimization,”Drones, vol. 3, no. 1, Jan. 2019

    2019

  9. [9]

    Survey on UA V cellular communications: Practical aspects, standardization advancements, regulation, and security chal- lenges,

    A. Fotouhiet al., “Survey on UA V cellular communications: Practical aspects, standardization advancements, regulation, and security chal- lenges,”IEEE Commun. Surv. Tutor ., vol. 21, no. 4, pp. 3417–3442, Mar. 2019

    2019

  10. [10]

    Communication and control in collaborative UA Vs: recent advances and future trends,

    S. Javaid, N. Saeed, Z. Qadir, H. Fahim, B. He, H. Song, and M. Bilal, “Communication and control in collaborative UA Vs: recent advances and future trends,”IEEE Trans. Intell. Trans. Sys., vol. 24, no. 6, pp. 5719– 5739, Jun. 2023

    2023

  11. [11]

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

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

    2022

  12. [12]

    Integrated sensing and communications framework for 6G networks,

    H. Luo, T. Zhang, C. Zhao, Y . Wang, B. Lin, Y . Jiang, D. Luo, and F. Gao, “Integrated sensing and communications framework for 6G networks,”IEEE Wireless Commun., vol. 32, no. 6, pp. 102–109, 2025

    2025

  13. [13]

    Toward 6G networks: Use cases and technologies,

    M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi, “Toward 6G networks: Use cases and technologies,”IEEE Commun. Mag., vol. 58, no. 3, pp. 55–61, Mar. 2020

    2020

  14. [14]

    Integrating sensing and communi- cations for ubiquitous IoT: Applications, trends, and challenges,

    Y . Cui, F. Liu, X. Jing, and J. Mu, “Integrating sensing and communi- cations for ubiquitous IoT: Applications, trends, and challenges,”IEEE Netw., vol. 35, no. 5, pp. 158–167, Nov. 2021

    2021

  15. [15]

    Future technology trends of terrestrial international mobile telecommunications systems towards 2030 and beyond,

    ITU, “Future technology trends of terrestrial international mobile telecommunications systems towards 2030 and beyond,” 2022

    2030

  16. [16]

    Reliable clutter suppression for slow-moving weak target radar detection,

    R. Zhang, J. Xue, and T. Zhang, “Reliable clutter suppression for slow-moving weak target radar detection,” in2025 IEEE International Conference on Communications Workshops (ICC Workshops), pp. 354– 359, 2025

    2025

  17. [17]

    Integrated sensing and communication enabled multiple base stations cooperative UA V detection,

    X. Lu, Z. Wei, R. Xu, L. Wang, B. Lu, and J. Piao, “Integrated sensing and communication enabled multiple base stations cooperative UA V detection,”arXiv e-prints, p. arXiv:2404.12705, Apr. 2024

    Pith/arXiv arXiv 2024

  18. [18]

    UA V trajectory monitoring for integrated sensing and communications system,

    S. Yan, H. Luo, P. Yang, J. Zhao, and F. Gao, “UA V trajectory monitoring for integrated sensing and communications system,”IEEE Trans. Wireless Commun., vol. 25, pp. 2733–2747, 2026

    2026

  19. [19]

    Symbol-level integrated sensing and communication enabled multiple base stations cooperative sensing,

    Z. Wei, R. Xu, Z. Feng, H. Wu, N. Zhang, W. Jiang, and X. Yang, “Symbol-level integrated sensing and communication enabled multiple base stations cooperative sensing,”IEEE Trans. V eh. Technol., vol. 73, no. 1, pp. 724–738, 2024

    2024

  20. [20]

    Learned off-grid imager for low-altitude economy with cooperative isac network,

    Y . Huang, J. Yang, S. Xia, C.-K. Wen, and S. Jin, “Learned off-grid imager for low-altitude economy with cooperative isac network,”IEEE Trans. Wireless Commun., vol. 25, pp. 3333–3348, 2026

    2026

  21. [21]

    Uav’s rotor micro-doppler feature extraction using integrated sensing and communication signal: Algorithm design and testbed evaluation,

    J. Wei, D. Ma, F. He, Q. Zhang, Z. Feng, Z. Liu, and T. Liang, “Uav’s rotor micro-doppler feature extraction using integrated sensing and communication signal: Algorithm design and testbed evaluation,” IEEE Trans. Wireless Commun., vol. 24, no. 12, pp. 10166–10182, 2025

    2025

  22. [22]

    Performance evaluation of micro-Doppler based UA V identification using different 5G frame structures,

    D. Maet al., “Performance evaluation of micro-Doppler based UA V identification using different 5G frame structures,” inProc. 2nd Int. Conf. Mobile Internet, Cloud Comput. Inf. Security, Changsha City, China, Apr. 2024, pp. 173–179

    2024

  23. [23]

    Dc-former network empowered uav and bird recognition based on integrated sensing and communication system,

    J. Xue, Q. Zhang, D. Ma, and J. Wei, “Dc-former network empowered uav and bird recognition based on integrated sensing and communication system,” in2025 10th International Conference on Computer and Communication System (ICCCS), pp. 927–932, 2025

    2025

  24. [24]

    Airguard: UA V and bird recognition scheme for integrated sensing and communications system,

    H. Luo, Z. Chu, T. Zhang, C. Zhao, B. Lin, and F. Gao, “Airguard: UA V and bird recognition scheme for integrated sensing and communications system,”IEEE J. Sel. Areas Commun., vol. 44, pp. 835–848, 2026

    2026

  25. [25]

    The 5th wireless big data competition, accessed: https://developer.huaweicloud.com/competition/information/1300000110,

    Huaweicloud, “The 5th wireless big data competition, accessed: https://developer.huaweicloud.com/competition/information/1300000110,” Available on March 13, 2026

    arXiv 2026

  26. [26]

    Integrated sensing and communication with mmWave massive MIMO: A compressed sampling perspective,

    Z. Gao, Z. Wan, D. Zheng, S. Tan, C. Masouros, D. W. K. Ng, and S. Chen, “Integrated sensing and communication with mmWave massive MIMO: A compressed sampling perspective,”IEEE Trans. Wireless Commun., vol. 22, no. 3, pp. 1745–1762, Mar. 2023

    2023

  27. [27]

    Integrated sensing and communications for V2I networks: Dynamic predictive beamforming for extended vehicle targets,

    Z. Du, F. Liu, W. Yuan, C. Masouros, Z. Zhang, S. Xia, and G. Caire, “Integrated sensing and communications for V2I networks: Dynamic predictive beamforming for extended vehicle targets,”IEEE Trans. Wireless Commun., vol. 22, no. 6, pp. 3612–3627, Jun. 2023

    2023

  28. [28]

    DoA estimation and capacity analysis for 3-D millimeter wave massive-MIMO/FD-MIMO OFDM systems,

    R. Shafin, L. Liu, J. Zhang, and Y .-C. Wu, “DoA estimation and capacity analysis for 3-D millimeter wave massive-MIMO/FD-MIMO OFDM systems,”IEEE Trans. Wireless Commun., vol. 15, no. 10, pp. 6963– 6978, Oct. 2016

    2016

  29. [29]

    Jittering effects analysis and beam training design for UA V millimeter wave communications,

    W. Wang and W. Zhang, “Jittering effects analysis and beam training design for UA V millimeter wave communications,”IEEE Trans. Wireless Commun., vol. 21, no. 5, pp. 3131–3146, Oct. 2022

    2022

  30. [30]

    Euler–rodrigues formula variations, quaternion conjugation and intrinsic connections,

    J. S. Dai, “Euler–rodrigues formula variations, quaternion conjugation and intrinsic connections,”Mech. Mach. Theory, vol. 92, pp. 144–152, Oct. 2015

    2015

  31. [31]

    MeshLab: an open-source mesh processing tool,

    P. Cignoniet al., “MeshLab: an open-source mesh processing tool,” in Proc. Eurographics Italian Chapter Conf., The Eurographics Associa- tion, Jul. 2008

    2008

  32. [32]

    The motion research laboratory, accessed: http://mocap.cs.cmu.edu,

    Carnegie Mellon University (CMU), “The motion research laboratory, accessed: http://mocap.cs.cmu.edu,” Available on March 13, 2026. Ac- cessed: http://mocap.cs.cmu.edu, 2026-03-13

    2026

  33. [33]

    Theory of communication. part 1: The analysis of informa- tion,

    D. Gabor, “Theory of communication. part 1: The analysis of informa- tion,”Journal of the Institution of Electrical Engineers-part III: radio and communication engineering, vol. 93, no. 26, pp. 429–441, 1946

    1946

  34. [34]

    A unified approach to short-time fourier analysis and synthesis,

    J. Allen and L. Rabiner, “A unified approach to short-time fourier analysis and synthesis,”Proceedings of the IEEE, vol. 65, no. 11, pp. 1558–1564, 1977

    1977

  35. [35]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” inProceedings of the IEEE/CVF international conference on computer vision, pp. 10012–10022, 2021

    2021

  36. [36]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  37. [37]

    A convnet for the 2020s,

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A convnet for the 2020s,” in2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 11966–11976, 2022

    2022

  38. [38]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, “An image is worth 16x16 words: Transformers for image recognition at scale,” 2021

    2021