A Joint Time-Varying Channel Estimation Based on Compressive Sensing and LSTM.

Xiaodong Han,Zihan Jiao,Peizhe Liang,Jiancun Fan

DOI: https://doi.org/10.1109/vtc2022-spring54318.2022.9860883

2022-01-01

Abstract:To achieve the theoretical performance gains of massive multiple-input multiple-output (MIMO) systems, the base station (BS) must acquire the downlink channel state information (CSI). In frequency division duplexing (FDD) massive MIMO systems, downlink CSI is estimated at user terminals with the pilot symbols transmitted by the BS at the first step and then user terminals feed it back to the BS. However, the huge number of antennas at the BS will result in heavy feedback overhead. Meanwhile, CSI acquisition is very challenging because of the high mobility of user terminals which causes the priori channel knowledge of the channels to change from one slot to another. In order to solve these problems, we propose a joint channel training and feedback scheme based on compressive sensing (CS) and deep learning (DL). Specifically, with the CS-based algorithm, named AS-JOMP, the sparse channel in time-delay domain can be adaptively reconstructed firstly. Then the DL-based network, named DnLSTM, is utilized to estimate the CSI. Simulation results demonstrate that the proposed method can reduce the training and feedback overhead and outperforms the existing classical algorithms at time-varying channel estimation.