Millimeter Wave Massive MIMO Channel Estimation Based On Bayesian Compressive Sensing

The ultra-high transmission rate and system capacity requirements of 5G wireless communication in the future promote the development of mobile networks,and the traditional microwave frequency is difficult to achieve this goal in the low frequency band,so it must be extended to the millimeter wave of the high frequency band.The physical properties of the millimeter wave determine its need to be combined with massive MIMO technology to provide a better user service experience.However,the use of large antenna arrays in millimeter-wave systems can increase system performance while incurring high overhead in system design.Compared to full-digital and full-analog architectures,the hybrid analog-to-digital architecture is a promising solution which can significantly reduce the number of RF links required for a system,and achieving a good balance between performance and power consumption.However,the signal processing problem under this system architecture has become a challenge,including channel modeling,channel estimation and channel precoding,etc.This paper first introduced the sparse distribution characteristics of millimeter-wave signals in the scattering space,and built a general mathematical model of the channel.Secondly,the different transceiver architectures and precoding schemes are analyzed and compared,and the power consumption model is used to evaluate the system architecture.Thirdly,the research on the reconstruction algorithm of sparse signal is carried out,and the application of Bayesian Compressive Sensing(BCS)theory in channel estimation is analyzed carefully.Then,for the problem of sparse channel solving in millimeter-wave massive MIMO systems,the problem of signal reconstruction is extended from the real domain to the complex domain by learning BCS inference analysis,and the channel estimation problem is transformed into a parameterized probability model,then the Expectation Maximization(EM)algorithm is utilized to solve the model parameters,which we call the EM-BCS algorithm.Different from the traditional convex optimization class and greedy iterative algorithm(such as Orthogonal Matching Pursuit(OMP)algorithm)for solving sparse channel estimation,EM-BCS algorithm can effectively solve the uncertainty problem in the system.Next,we constructed a millimeter-wave massive MIMO-OFDM system model based on the fully-connected architecture,which considered the horizontal direction angles of both sides of the transceiver,and used a uniform linear array antenna structure.The EM-BCS algorithm was applied to the channel estimation of the system and numerical simulation was performed.The experimental results showed that compared with the channel estimation accuracy based on OMP algorithm,the channel estimation based on EM-BCS algorithm has better reconstruction performance.Finally,we constructed a multi-user millimeter-wave massive MIMO system model based on the sub-connection architecture,which considered the elevation angles outside the horizontal direction angles,and used a uniform planar array antenna structure.The experimental results showed that the EM-BCS algorithm can realize the application expansion from single-user to multi-user,and the spectrum efficiency of multi-user system is significantly better than that of the general single-user system.