Massive MIMO Channel Estimation And Channel Prediction Methods Based On Measured Channels

Massive MIMO technology can significantly improve the system's spectrum utilization efficiency and data transmission rate by deploying a large-scale antenna array at the base station to serve multiple users at the same time.The performance of a massive MIMO system is heavily dependent on CSI,so accurate CSI acquisition is the key to system performance improvement.Due to the fast-changing and non-stationary characteristics of the channel in massive MIMO systems,the problem of outdated CSI will occur when using traditional channel estimation methods.Aiming at the problem of how to obtain accurate CSI in the massive MIMO system,this paper conducts an in-depth study on the channel estimation method and channel prediction method in the massive MIMO system.The main research contents of this paper are as follows:First,through time-domain autocorrelation analysis and frequency-domain autocorrelation analysis of the measured channel data,it is found that the time autocorrelation of the CSI of the measured channel is poor,but the autocorrelation in the frequency domain is strong.Secondly,according to the strong autocorrelation characteristics of the measured channel data in the frequency domain,a pilot-assisted channel estimation method based on deep convolutional residual network(DCRes Net)is proposed.This method only needs to use a small number of pilot signals in the frequency domain,and the CSI on all frequency points can be recovered through the interpolation method,and high channel estimation accuracy can be guaranteed.Compared with the existing interpolation-based channel estimation methods,under the premise of achieving the same channel estimation accuracy,the proposed method uses significantly less pilots than the existing methods,which effectively improves the spectrum utilization efficiency of the system and data transfer rate.Finally,according to the characteristics of weak time domain correlation and strong frequency domain correlation of the measured data,a time-frequency combined channel prediction method based on convolution long short-term memory(Convolution long shortterm memory,Conv LSTM)is proposed.The method combines the characteristics of the time domain and the frequency domain to predict the channel,and uses the strong autocorrelation of the frequency domain to improve the accuracy of the time domain prediction.Compared with the existing methods that only use time-domain correlation for channel prediction,the proposed method has higher channel prediction accuracy.