EUVIMEDEUVIMED entdecken
Crossref

Fundamental Limits of Gaussian-Mixture CSI Compression for Massive MIMO

Bumsu Park, Youngmok Park, Chanho Park, Namyoon Lee

IEEE Transactions on Wireless Communications · 2026 · 25 · 25580-25596

DOI: 10.1109/twc.2026.3746604 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Wireless Communications, 2026. Verfasst von Bumsu Park und 3 weiteren Autor:innen. Die Arbeit nennt 64 Literaturangaben. Gefördert von: National Research Foundation of Korea (NRF) Grant funded by Korean Government, Institute of Information and Communications Technology Planning and Evaluation (IITP) Grant funded by Korean Government, Korean ARPA-H Project through Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health and Welfare, Republic of Korea.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Channel state information (CSI) feedback in frequency-division duplex (FDD) massive multiple-input multiple-output (MIMO) systems is fundamentally limited by the high dimensionality of wideband channels. In this paper, we model the stacked wideband CSI vector as a Gaussian-mixture source with a latent geometry state: each component is a locally stationary propagation regime with its own covariance, capturing the multimodality of practical CSI datasets while preserving analytical tractability. Building on this model, we propose Gaussian-mixture transform coding (GMTC), a practical CSI feedback architecture in which the user equipment infers the most likely geometry state, encodes its label losslessly, and compresses the CSI with the Karhunen–Loéve transform matched to that state, using a mixture dictionary learned offline and shared with the base station. For a finite Gaussian-mixture source, whose exact rate–distortion (RD) function is generally unknown, we derive a genie-aided lower bound and a true-label achievable upper bound. The gap between these bounds is at most the amortized state-label entropy and therefore vanishes when the mixture order grows subexponentially with the CSI dimension. A key structural result reveals that the optimal rate allocation across all mixture components is governed by a single global reverse-waterfilling level. We also quantify the cost of the idealizations underlying these bounds: an RD-stability result controls the deviation from the RD function of the true CSI source in terms of the Wasserstein-2 modeling error, while a closed-form bound on the excess distortion incurred by practical maximum-a-posteriori state selection decays exponentially with the CSI dimension. Simulations on COST2100 and the ray-tracing-based DeepMIMO datasets show that GMTC significantly improves the RD tradeoff over nonlinear transform coding approaches at substantially smaller memory and lower complexity.

Förderung

  • National Research Foundation of Korea (NRF) Grant funded by Korean Government
  • Institute of Information and Communications Technology Planning and Evaluation (IITP) Grant funded by Korean Government
  • Korean ARPA-H Project through Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health and Welfare, Republic of Korea

Zitieren

APA

Park, B., Park, Y., Park, C., & Lee, N. (2026). Fundamental Limits of Gaussian-Mixture CSI Compression for Massive MIMO. IEEE Transactions on Wireless Communications, 25, 25580-25596. https://doi.org/10.1109/twc.2026.3746604

Vancouver

Park B, Park Y, Park C, Lee N. Fundamental Limits of Gaussian-Mixture CSI Compression for Massive MIMO. IEEE Transactions on Wireless Communications. 2026;25:25580-25596. doi:10.1109/twc.2026.3746604

BibTeX
@article{Park2026,
  author = {Park, Bumsu and Park, Youngmok and Park, Chanho and Lee, Namyoon},
  title = {Fundamental Limits of Gaussian-Mixture CSI Compression for Massive MIMO},
  journal = {IEEE Transactions on Wireless Communications},
  year = {2026},
  volume = {25},
  pages = {25580--25596},
  doi = {10.1109/twc.2026.3746604}
}
Kostenlos starten →