We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.IT

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Information Theory

Title: Massive MIMO Channel Prediction Using Machine Learning: Power of Domain Transformation

Abstract: To compensate the loss from outdated channel state information in wideband massive multiple-input multipleoutput (MIMO) systems, channel prediction can be performed by leveraging the temporal correlation of wireless channels. Machine learning (ML)-based channel predictors for massive MIMO systems were designed recently; however, the time overhead to collect a large amount of training data directly affects the latency of the system. In this paper, we propose a novel ML-based channel prediction technique, which can reduce the time overhead to collect the training data by transforming the domain of channels from subcarrier to antenna in wideband massive MIMO systems. Numerical results show that the proposed technique can not only reduce the time overhead but also give additional performance gain compared to the ML-based channel prediction techniques without the domain transformation.
Subjects: Information Theory (cs.IT); Signal Processing (eess.SP)
Cite as: arXiv:2208.04545 [cs.IT]
  (or arXiv:2208.04545v1 [cs.IT] for this version)

Submission history

From: Beomsoo Ko [view email]
[v1] Tue, 9 Aug 2022 05:20:15 GMT (1232kb)

Link back to: arXiv, form interface, contact.