Joint Angular- and Delay-Domain MIMO Propagation Parameter Estimation Using Approximate ML Method
Joint Angular- and Delay-Domain MIMO Propagation Parameter Estimation Using Approximate ML Method
复制标题
使用近似 ML 方法的联合角域和延迟域 MIMO 传播参数估计
DOI：
10.1109/tsp.2007.896247
复制
发表时间：
2007
影响因子：
5.4
通讯作者：
V. Koivunen
中科院分区：
文献类型：
--
作者：
C. Ribeiro;A. Richter;V. Koivunen
本文提出了一种联合估计多输入多输出（MIMO）信道测深中经常观测到的集中传播路径和分布散射分量参数的估计方法。联合角延迟域模型导致了一个高维的相关矩阵，使得最大似然估计无法直接实现。我们推导了利用协方差矩阵的结构计算近似ML估计的低复杂度方法。我们提出了一个迭代的两步程序，交替估计集中传播路径的参数和分布散射的参数。对于分布式散射，该估计器首先优化描述其时延结构的参数。然后，利用估计的时滞参数，对角分布参数进行优化。我们给出了仿真结果，并将估计的时延和角分布与实际分布进行了比较，表明获得了高质量的估计。通过建立Cramer-Rao下界（CRLB）并将其与估计的方差进行比较，研究了估计器的大样本性能。仿真结果表明，对于大多数参数，在相对较小的样本量下，所提出的估计方法的方差达到了CRLB，并且没有观察到偏差。
In this paper, we derive an estimation method that jointly estimates the parameters of the concentrated propagation paths and the distributed scattering component that are frequently observed in multiple-input multiple-output (MIMO) channel sounding measurements. The joint angular-delay domain model leads to a correlation matrix with high dimensionality, which makes direct implementation of a maximum-likelihood (ML) estimator unfeasible. We derive low-complexity methods for computing approximate ML estimates that exploit the structure of the covariance matrices. We propose an iterative two-step procedure that alternates between the estimation of the parameters of the concentrated propagation paths and the parameters of the distributed scattering. For the distributed scattering, the estimator first optimizes the parameters describing their time-delay structure. Then, using the estimated time-delay parameters, the parameters of the angular distributions are optimized. We present simulation results and compare the estimated time-delay and angular distributions to the actual distributions, demonstrating that high-quality estimates are obtained. The large sample performance of the estimator is studied by establishing the Cramer-Rao lower bound (CRLB) and comparing it to the variances of the estimates. The simulations show that the variance of the proposed estimation technique reaches the CRLB for relatively small sample size for most parameters, and no bias is observed.