A Variational Bayesian Approach for Channel Estimation in Pilot-Contaminated User-Centric C-RAN System

• Published in: IEEE access Vol. 12; p. 1
• Main Authors: Bera, Soumyasree, Vinnakota, Venu Balaji, Sen, Debarati
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: 5G mobile communication; Algorithms; Bayes methods; Bayesian analysis; Bit error rate; Channel estimation; Cloud radio access networks; Clusters; Cramer-Rao bounds; Estimation; Iterative methods; iterative variational Bayesian inference; millimeter wave; Millimeter wave communication; Performance degradation; pilot contamination; Radio frequency; Statistical inference; Training; User experience; User-centric cloud radio access network; Vectors; Wireless communication systems
• DOI: 10.1109/ACCESS.2024.3395170
• ISSN: 2169-3536

In fifth-generation and beyond (5GB) wireless communication systems, Cloud Radio Access Network (C-RAN) is recognized as a vital technology. In this endeavor, User-Centric C-RAN (UC-RAN) promises a significant reduction in the channel training overhead because only the intra-cluster channel state information (CSI) is necessary for successful high data rate transmission. However, the network performance may be degraded by inter-cluster interference. Furthermore, such networks are susceptible to the pilot contamination effect, which leads to a major restriction on overall system performance. To tackle this issue, we introduce a channel estimation (CE) approach based on iterative variational Bayesian inference (IVBI)-called the least square IVBI (L-IVBI) scheme. This method consists of two stages: initialization and iteration, and is designed for the UC-RAN system. The initialization stage includes a coarse channel estimate, further refined in the iteration stage. We follow the alternative minimization method to estimate the desired channel in the iteration stage. Extensive simulation results for the UC-RAN system validate the proposed algorithms. We also provide the derivation of the Bayesian Cramer-Rao bound (BCRB) for the proposed estimator. The novel approach significantly outperforms the state-of-the-art in terms of normalized mean square error (NMSE), spectral efficiency (SE), and bit error rate (BER).