Research On Channel Estimation Of Low Precision Quantized Massive MIMO System

With the rapid growth of the bandwidth and data transmission rate of wireless communication systems,future mobile communication will face the problems of high hardware cost and system power consumption,which will become the basic bottleneck restricting the substantial improvement of system performance.In order to solve the above problems,the academic community has launched research on low-power and high-efficiency wireless transmission technologies,in which receivers based on lowprecision(such as 1-3 bit)ADCs can effectively avoid the high cost and high cost of high-precision ADCs.The problem of power consumption.However,the channel estimation algorithm suitable for highprecision quantization systems is not suitable for low-precision quantization systems.Therefore,this paper mainly studies the channel estimation problem of one-bit quantization massive MIMO systems.The main research content and innovation points of this paper can be divided into the following two parts:(1)First,two channel estimation algorithms suitable for massive MIMO systems with one ADC are proposed.The paper first gives an ADC large-scale MIMO system model,and then iteratively optimizes the maximum likelihood algorithm and the low-rank algorithm through the optimization minimization theory,that is,the maximum likelihood channel estimation algorithm(M-ML)based on optimization minimization And the optimal minimization-based low-rank channel estimation algorithm(M-LR).The M-ML algorithm works by finding a simple optimization function for the one-bit log-likelihood function,which can iteratively update the analytical solution of the channel matrix and noise variance estimation.Considering the low-rank characteristics of the mmWave massive MIMO channel matrix,the M-LR algorithm is proposed,that is,a kernel-normal penalty term based on the channel matrix is added to the one-bit negative logarithmic likelihood function.Finally,the computational complexity of the two algorithms is analyzed,and the experiments verify that the M-LR algorithm is superior to the M-ML algorithm in terms of performance and iteration speed.Under low signal-to-noise ratio,compared with the M-ML algorithm,using the low-rank feature can provide a performance gain close to 14dB.The performance changes of the two algorithms are also compared when the regularization parameters are updated adaptively or when the regularization parameters are fixed.(2)Secondly,considering the angle-domain channel model of a massive MIMO system with one ADC in the millimeter-wave scenario,two angle-domain parametric channel estimation algorithms with relatively high computational efficiency are proposed,which are relaxation-based maximum likelihood channel estimation.algorithm(R-ML)and relaxation-based low-rank channel estimation algorithm(RLR).The R-ML algorithm uses a relaxation-based loop algorithm to obtain maximum likelihood estimation,and it is found through analysis that its computational complexity is relatively high.In order to solve this problem,the R-LR algorithm was then proposed.Finally,the computational complexity of the two algorithms is analyzed,the performance and iteration speed of the two algorithms are compared through experiments,and the performance and iteration speed of the R-LR algorithm are verified to be superior to the R-ML algorithm.The iteration speed of the R-LR algorithm is About 30 times faster than the R-ML algorithm.It is considered to introduce Bayesian Criterion(BIC)into the parameter channel estimation algorithm to estimate the number of channel paths.Finally,the accuracy of using Bayesian criterion to determine the number of paths is verified by experiments.