Research On Large-scale MIMO Channel Estimation And Detection Method Based On Deep Network

The innovation of wireless communication technology has brought great convenience to people’s social life,and as the deployment of 5G becomes more and more perfect,the research about 6G also begins to start gradually.MIMO signal detection and MIMO-OFDM channel estimation is an important branch of wireless communication physical layer technology and a key technology in 5G and 6G,so it has attracted a lot of attentions from researchers and engineers.At present,deep learning techniques have been successfully applied in many fields and achieved excellent results,and the application of deep learning to the physical layer has become an important research branch in the field of wireless communication.In this paper,we focus on two problems,MIMO signal detection and MIMO-OFDM channel estimation,and propose new deep learning-based algorithms based on previous work.In MIMO signal detection,this paper analyzes and optimizes the structure of existing MIMO detection algorithms based on deep learning,and thus proposes a new algorithm with better mis-symbol rate performance,fewer parameters and lower complexity.Simulation results show that in a scenario where the receiving antenna is twice as many as the transmitting antenna,the false symbol rate performance of our proposed algorithm is close to that of the sphere decoding algorithm for both low-order QPSK modulation and high-order 16QAM modulation.In terms of MIMO-OFDM channel estimation,this paper considers the existence of channel frequency domain correlation and space domain correlation in the system,and proposes a convolutional neural network-based MIMOOFDM channel estimation algorithm for large-scale MIMO scenarios by analyzing the feasibility of convolutional layers in interpolation.Simulation results show that the NMSE performance of the proposed algorithm is superior to that of the traditional estimation algorithm combining least squares and linear interpolation.