Large-scale MIMO channel state information feedback method based on deep learning

Abstract

The invention relates to a large-scale MIMO channel state information feedback method based on deep learning, and the method comprises the steps: carrying out the two-dimensional discrete Fourier transformation of a channel matrix of MIMO channel state information in a space-frequency domain at a user side in a downlink, and obtaining a sparse channel matrix H in an angle delay domain; a super-resolution channel state information network model is constructed, the model comprises a decoder belonging to a user side and a decoder belonging to a base station side, and the decoder is used for coding the obtained sparse channel matrix H in the angle delay domain into lower-dimension code words; the decoder is used for reconstructing a channel matrix estimation value from received code words to train the super-resolution channel state information network model, so that the channel matrix estimation value is close to a channel matrix H sparse in an angle delay domain as much as possible, and model parameters are obtained; and applying the trained super-resolution channel state information network model to compressed sensing and reconstruction of channel information.

Description

technical field [0001] The present invention relates to the field of communication technologies, in particular, to a massive MIMO channel state information feedback method based on deep learning. Background technique [0002] Massive MIMO (Multiple-Input Multiple-Output) technology is considered to be one of the three core technologies of 5G. Massive MIMO technology has spatial multiplexing gain, diversity gain and beamforming capabilities. By configuring multiple antennas at the transmitter and receiver to achieve multiple reception and multiple transmission, it can make full use of space resources, without increasing spectrum resources and transmit power. , which doubles the channel capacity and reduces multi-user interference, showing significant performance advantages, but these gains are based on the premise that the base station can accurately know the uplink and downlink channel state information (CSI, channelstate information). Uplink CSI acquisition is relatively e...

AI Technical Summary

Problems solved by technology

The above-mentioned compressive sensing-based methods are relatively advanced channel feedback methods at present, but there are still the following problems: 1) Compressive sensing algorithms generally rely on prior assumptions about the channel structure, that is, the channel satisfies sparsity on a certain transformation basis, and In practice, the channel is not completely sparse in any transformation basis, and has a more complex structure, so the algorithm based on compressed sensing relies on more complex prior conditions; 2) compressed sensing uses random projection methods to obtain low-dimensional compressed signals, so The complete channel structure has not been obtained; 3) Most of the existing compressed sensing algorithms are iterative algorithms, which require huge computing overhead and pose a huge challenge to the real-time performance of the system

[0004] Publication No. CN108390706A, patent document titled "A Method for Feedback of Large-Scale MIMO Channel State Information Based on Deep Learning" introduces deep learning technology into the large-scale MIMO feedback scheme, providing a new design for solving the CSI feedback problem under the FDD system However, this scheme still faces low CSI reconstruction accuracy and cannot preserve the channel to the greatest extent in low feedback overhead and complex outdoor scenes. complete information question

Images