基于改进EM算法的分布式MIMO系统联合参数估计

Joint parameter estimation for distributed MIMO system based on improved EM algorithm

鉴于采用最大似然算法估计分布式多天线系统的信道增益与频偏存在多维优化使计算复杂度高的缺陷,以及采用期望最大化（EM）算法存在收敛速度慢、对初值依赖性大的不足,而期望条件最大化（ECM）算法用一系列计算简单的CM步来代替一个复杂的M步,弱化了初值对收敛性的影响.综合考虑ECM算法与最大似然（ML）估计算法来优化EM的收敛过程,提出了一种高效稳定的EM算法.该算法在CM步取得频偏的更新值后,通过ML的结果来更新信道增益.仿真结果表明：该算法对初值的依赖性较低、计算简单且稳定性高,结合空间选择期望最大化（SAGE）方法后能大幅提高收敛速度,且所得估计值的均方误差（MSE）能够逼近Cramer-Rao界（CRB）.

In consideration of the drawback of high computational complexity related to multi-dimentional optimization of the maximum-likelihood algorithm to estimate the channel gain and the frequency offset for distributed MIMO;system,and the problems of slow convergence rate of the Expectation-Maximization（EM）algorithm and its dependence on the initialization values,the Expectation Conditional Maximization（ECM）Algorithm reduces the effect of the initialization values with a series of easy computing CM step instead of a complicated M step.This paper optimizes the convergence process of EM algorithm by combining the ECM algorithm and the ML algorithm simultaneously,after acquiring the latest value of frequency offset at CM step to update the channel gain by the result of ML,and presents a EM type algorithm which has high efficient and stability.From simulation results we can see the presented algorithm has low sensitive to initialization values,simply computational and high stability properties.Combining with the Space-alternating generalized expectation maximization（SAGE）algorithm the algorithm can improve the convergence rate significantly and the mean-square-error（MSE）can approach the Cramer-Rao bound.