Proposing LMU-GRUs Based Channel Estimation in Massive MIMO System

Abstract

A Massive Multiple-Input Multiple-Output (massive MIMO) system relies on channel state information (CSI) feedback to perform precoding and achieve performance gain in frequency division duplex (FDD) networks. However, the transmission of a massive MIMO system is subject to excessive feedback overhead. In this paper, we propose a Deep Learning (DL) approach-based channel estimation technique to enhance recovery quality and improve the trade-off between compression ratio (CR) and complexity of a massive MIMO system. The first proposed technique uses the Channel State Information Network combined with the gated recurrent unit (CsiNet-GRU). The second proposed technique is based upon using CsiNet combined with Legendre Memory Unit in GRUs layers (CsiNet-LMU-GRUs) which improves a novel memory cell for recurrent neural networks that dynamically maintains along by information across long windows of time using relatively few resources. Moreover, the proposed techniques use the dropout method to reduce overfitting during the learning process. The simulation results demonstrate that the proposed (CsiNet-GRU) and (CsiNet-LMU-GRUs) techniques result in a significant performance improvement when compared with existing techniques used in conjunction with massive MIMO systems.