太赫兹MIMO系统中基于SRCGAN的空时频信道估计方案OA

Terahertz's input multi-output technology is of bandwidth and high reuse gain,which can provide higher data rates and larger network capacity.It is the core technology of the sixth generation of mobile communication.However,the severe frequency selective decline effect and a large number of antennas in the Terahertz MIMO system have brought huge challenges to channel estimates.Traditional channel estimation methods are difficult to accurately estimate channels in the increasingly complex wireless communication environment.In the past research,some scholars combine super-resolution with deep learning network to propose the channel estimation framework.The network estimates all channel status information based on the channel information of the pilot.However,the previous work was mainly concentrated in the mapping of learning frequency domains and time domain channels,and there was no comprehensive consideration of the channel correlation among time,space and frequency domain.This article focuses on the three-dimensional channel correlation and proposes a channel estimation algorithm based on super resolution conditional generative adversarial networks.In this scheme,the initial space-time channel response matrix in this scheme is regarded as a two-dimensional low-resolution picture.Then,the space-time characteristics of the terahertz channel are extracted by the super resolution generative adversarial network,and the high-resolution picture is obtained by the adversarial learning of the generator and discriminator,namely the complete channel matrix.In order to make full use of the multi-dimensional channel characteristics of terahertz channel and further improve the performance of the channel estimation,this thesis proposes a channel estimation algorithm based on super resolution conditional generative adversarial network.The data is optimized with pre-arranged methods and the structure of deep neural networks.In this scheme,we cache the channel matrix of the previous subcarrier based on the rough estimate of pilot frequency and input it into the current super resolution conditional generative adversarial network as a conditional label.By introducing additional prior information to constrain the generator and discriminator,the generated adjoint network is extended into a conditional model and the generator is guided to generate.What's more,when the distribution of channel characteristics changes,the network can adapt to learn the change of distribution,thus increasing the robustness of the system.At the same time,the online prediction module based on gradient descent is used to improve the robustness of the system.Simulation results show that the application of the image super resolution in the channel estimation problem can effectively improve the accuracy of the estimation results.The super resolution generative adversarial network shows good channel estimation performance because of its superior capability of global feature extraction and space-time correlation extraction.The performance of the proposed adversarial network based on super-resolution conditions is better than that of traditional interpolation algorithms and super-resolution convolutional networks.At the same time,the information provided by the frequency correlation between adjacent subcarriers can more effectively utilize the"time,frequency and space"three-dimensional correlation of channel,and its performance is more outstanding than that of the space-time channel estimation scheme based on super resolution generative adversarial network.At the same time,the improved online prediction module is used to ensure the robustness of the network.