DBPN Based Uplink Channel Estimation for Multi-User MISO RIS System

In this letter, we propose two uplink channel estimation methods based on deep back projection networks (DBPN), one of the deep learning models for super-resolution, for reconfigurable intelligent surfaces (RIS) systems. The first method is to estimate the entire channel with partial channel samples...

Full description

Saved in:
Bibliographic Details
Published in:IEEE wireless communications letters 2023-12, Vol.12 (12), p.1-1
Main Authors: Seo, Jeongbin, Choi, Geonho, Kim, Suk Chan
Format: Article
Language:English
Subjects:
Channel estimation
Convolution
Deep learning
Estimation
Reconfigurable intelligent surfaces (RIS)
Super-resolution
Superresolution
Symbols
Uplink
Uplink channel estimation
Uplinking
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this letter, we propose two uplink channel estimation methods based on deep back projection networks (DBPN), one of the deep learning models for super-resolution, for reconfigurable intelligent surfaces (RIS) systems. The first method is to estimate the entire channel with partial channel samples using the DBPN. Due to the iterative up and down projection architecture of the DBPN, the channel can be accurately estimated with relatively small samples. However, the DBPN requires high computational complexity in the process of transferring information between layers. To solve this problem, we modify the model architecture. The second method is a channel estimation method using the simplified DBPN. Through simulations, it is confirmed that the proposed methods have superior channel estimation performance than benchmark schemes. Furthermore the proposed method with simplified DBPN has lower complexity than the method using the original DBPN without performance degradation. In this letter, we propose two uplink channel estimation methods based on deep back projection networks (DBPN) for reconfigurable intelligent surfaces (RIS) systems. The proposed method is to estimate the entire channel with partial channel samples using the DBPN. Due to the iterative up and down projection architecture of the DBPN, the channel can be accurately estimated with relatively small samples. However, the DBPN requires high complexity in transferring information between layers. To solve this problem, we also propose the channel estimation method with the simplified DBPN. Simulation results show the superior performances of the proposed methods than benchmark schemes.
ISSN:2162-2337
2162-2345
DOI:10.1109/LWC.2023.3309844