高斯混合噪声中弱信号的Rao检测方法  

The Rao Detection of Weak Signal in Gaussian Mixture Noise

作　　者：方前学[1,2] 王永良[3] 王首勇[2] 

机构地区：[1]国防科技大学电子科学与工程学院,长沙410073 [2]空军雷达学院重点实验室,武汉430019 [3]空军雷达学院科研部,武汉430019

出　　处：《信号处理》2009年第5期751-754,共4页Journal of Signal Processing

摘　　要：期望最大化(Expectation Maximization,EM)算法是求解参数最大似然估计(MLE)的最优迭代算法,但若参数初始化不恰当,会使估计值落入"初值陷阱",导致错误的参数估计值。为此,本文提出了估计高斯混合噪声参数的矩-EM算法,即先求参数的矩估计,并用矩估计值初始化参数,再通过EM迭代算法估计参数。在此基础上,经高斯化滤波,导出了高斯混合噪声背景下未知幅度弱信号的Rao检验统计量。仿真结果表明,矩-EM算法可以更准确地估计噪声参数;基于矩-EM算法的Rao检测性能优于基于EM算法的Rao检测性能。The expectation maximization （EM） algorithm is available for computing the maximum likelihood estimation （MLE） of parameter. However, if there is not an appropriate initialization, the ite＇rative computation will stop at initialization trap and lead to im- proper estimation. In this paper,the moment-EM algorithm is proposed to overcome the problem. It means to compute the moment-estima- tion of parameter and initialize parameter with moment-estimation firstly, and then amend the estimation through EM algorithm. In suc- cession, with the Gaussian filter based on estimated parameter, the paper presents the Rao decision rule of the weak signal with unknown amplitude under Gaussian mixture noise environment. Simulation results indicate that moment-EM algorithm can estimate parameter accurately and the detection performance of Rao test based on moment-EM Algorithm outperforms that of Rao test based on EM Algorithm.

关 键 词：高斯混合噪声 参数估计 矩-EM算法 Rao检测 

分 类 号：TN957.51[电子电信—信号与信息处理]