An attention-based channel estimation algorithm for next-generation point to point communication systems

Kayode A. Olaniyi, Reolyn Heymann, Theo G. Swart

Research output: Contribution to journalArticlepeer-review

Abstract

Accurate and robust estimation of channel parameters is essential in establishing reliable communication with characteristic optimal resource utilization in next-generation communication systems. Traditional techniques have limitations, such as the need for additional bandwidth and decreased spectral efficiency. Thus, there is a need for novel techniques that enhance the accuracy and robustness of channel parameter estimation in next-generation communication systems. To address this need, we propose in this paper a recurrent neural network (RNN)-based attention mechanism, to improve channel estimation accuracy and robustness in next-generation communication systems. The attention mechanism selectively focuses on the most relevant features while ignoring noise and interference. The attention network weights are initialized and are constantly updated in the course of network training. The weight values determine the significance of the features before passing them to the channel estimator. This allows the algorithm to adapt to varying channel conditions and improve its accuracy in challenging environments. The proposed attention-based algorithm performance is compared with three baseline techniques: learned denoising-based approximate message passing (LDAMP), Wasserstein generative adversarial networks (WGAN), and maximum likelihood (ML). The result evaluations indicate that the attention-based algorithm performs better than the existing artificial intelligence-based channel coding algorithms, in terms of robustness and accuracy.

Original languageEnglish
Pages (from-to)4018-4029
Number of pages12
JournalBulletin of Electrical Engineering and Informatics
Volume13
Issue number6
DOIs
Publication statusPublished - Dec 2024

Keywords

  • Attention mechanism
  • Channel estimation
  • Channel matrix
  • Encoder-decoder
  • Pilot densities

ASJC Scopus subject areas

  • Computer Science (miscellaneous)
  • Control and Systems Engineering
  • Information Systems
  • Instrumentation
  • Hardware and Architecture
  • Computer Networks and Communications
  • Control and Optimization
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'An attention-based channel estimation algorithm for next-generation point to point communication systems'. Together they form a unique fingerprint.

Cite this