Deep Learning based Channel Estimation and Hybrid Beamforming for 5G Massive MIMO Wireless Communications

Abstract

Hybrid beamforming (BF), which divides beamforming operation into radio frequency (RF) and baseband (BB) domains, will play a critical role in MIMO communication at millimeter-wave(mmWave) frequencies. This paper also introduce offline training and prediction schemes for channel estimation and hybrid beamforming. The aim of this paper is that to increase spectral efficiency over more data streams by leveraging the deep learning based LSTM network. The LSTM network is used to train the numeric values from sequence data and predict on new sequence data. The performance is evaluated under different parameters including number of data streams (1, 2, 3 and 4) with different signal-to-noise ratio (SNR) for different carrier frequencies (28GHz, 38GHz, 60GHz and 73GHz) through computer simulation using MATLAB. The simulation results verified that the proposed method can achieve higher spectral efficiency when the number of data streams increases and the value of SNR-Test increases too.