一种基于EM算法的快速收敛参数估计方法  

Fast convergence parameter estimation method based on expectation-maximum algorithm

作　　者：王戈[1] 于宏毅[1] 沈智翔[1] 胡赟鹏[1] 

机构地区：[1]解放军信息工程大学信息工程学院,郑州450002

出　　处：《吉林大学学报（工学版）》2013年第2期532-537,共6页Journal of Jilin University:Engineering and Technology Edition

基　　金：'973'国家重点基础研究发展规划项目(613148);国家科技重大专项项目(2008ZX03006)

摘　　要：将EM算法用于参数估计中,提出了一种在EM算法迭代中使用符号后验概率修正先验概率的快速收敛参数估计方法。通过分析参数估计的CRB与EM算法收敛速率的关系,指出通过降低参数估计的CRB可以提高EM算法的收敛速率。证明了修正之后的算法能加速算法收敛的机理,即降低了缺失数据的熵;同时证明了修正后的算法仍然收敛到修正前的似然函数。最后以载波相位估计为例与传统基于EM算法的相位估计方法进行比较,仿真结果表明,在不影响估计性能的前提下,算法收敛速率明显加快。A fast convergence parameter estimation method based on Expectation-Maximum （EM） algorithm is developed. This method modifies the priori probability with posterior1 probamnty m me iteration of the EM algorithm. The relationship between the CRB of parameter estimation and the convergence speed of the EM algorithm is analyzed. It is shown that to decrease the CRB of parameter estimation can accelerate she convergence speed of the EM algorithm. The mechanism to accelerate the convergence speed by the modified algorithm is proved, that is to reduce the entropy of the missing data. It is also proved that the modified algorithm can converge to the likelihood function as the non- modified algorithm. Taking phase estimation as example, the modified method is compared with traditional method based on EM algorithm. Simulation results show that, without affecting the estimation performance, the convergence speed of the modified method is faster.

关 键 词：通信技术 期望最大化算法 先验概率 收敛速率 同步参数估计 CRAMER-RAO下界 

分 类 号：TN911.23[电子电信—通信与信息系统]