Channel prediction method and system based on joint time-frequency correlation for large-scale mimo systems

Abstract

The invention discloses a channel prediction method and system based on time-frequency joint correlation of a massive MIMO system, and belongs to the field of massive MIMO wireless communication. According to the characteristics of weak time-domain correlation and strong frequency-domain correlation of measured data, the present invention proposes a time-frequency combined channel prediction method based on convolutional long-term short-term memory network. Convolutional LSTM is a method that can simultaneously extract time domain and frequency domain. The deep learning model of domain features, which extracts the frequency domain features of the channel through the input convolution structure, and uses the internal LSTM structure to extract the time domain features of the channel, and applies the channel features in the frequency domain to the channel prediction in the time domain. In this way, the effect of time-frequency joint channel prediction is achieved. This method combines the characteristics of time domain and frequency domain to predict the channel, and uses the strong autocorrelation of frequency domain to improve the accuracy of time domain prediction. Compared with existing channel prediction methods that only use time-domain correlation, the proposed method has higher channel prediction accuracy.

Description

Technical field [0001] The present invention belongs to the field of large-scale multi-input multi-output ( MIMO) wireless communication, more particularly, relates to a large-scale MIMO system based on a channel prediction method and system of time-frequency joint correlation. Background [0002] As a key technology in 5G, massive MIMO technology has been widely studied in recent years. Massive MIMO technology achieves greater spatial diversity gain by placing more antennas, usually hundreds or thousands, at the base station (BS) end, while serving multiple users, further improving data transmission rates and link reliability. In order to take full advantage of large-scale MIMO systems, the acquisition of accurate channel state information (CSI) is essential. CSI estimation is generally carried out by the method of conduction estimation, but in large-scale MIMO systems, this method of conduction estimation will have the problem of CSI obsolescence. At present, the channel predic...

AI Technical Summary

Problems solved by technology

For traditional channel prediction methods based on parameter models, due to the increase in the number of antennas, the computational complexity required for parameter estimation also increases

For the AR model, although its computational complexity is much smaller than that of parameter-based channel estimation, it models the change of the channel as an autoregressive process, only considering the correlation of the channel in time

However, according to the exploration of measured channel data, it is found that in massive MIMO systems, the CSI in the channel changes rapidly with time, and the temporal correlation is low. It is difficult to predict the channel simply by using the temporal correlation. achieve the desired effect

[0004] To sum up, the existing channel prediction schemes may no longer be suitable for massive MIMO systems, so the present invention proposes a channel prediction method based on time-frequency joint correlation, which combines time-domain and frequency-domain correlation to Predict the channel and use the high autocorrelation in the frequency domain to improve the accuracy of time domain prediction

Images