α稳定分布特征指数估计方法比较  

Comparison of Estimation Methods for Characteristic Exponent ofα-Stable Distribution

作　　者：刘文红[1] 张帅[2] 邱天爽[3] 

机构地区：[1]上海电机学院电子信息学院,上海200240 [2]上海电机学院电气学院,上海200240 [3]大连理工大学电子信息与电气工程学部,辽宁大连116024

出　　处：《上海电机学院学报》2013年第4期183-187,共5页Journal of Shanghai Dianji University

基　　金：国家自然科学基金资助(61172108);上海市教育委员会科研创新项目资助(12YZ187);上海电机学院重点学科资助(10XYF01);上海电机学院横向合作项目资助(12B27)

摘　　要：α稳定分布通常用其特征函数来描述。在特征函数的4个参数中,特征指数α是最重要的一个参数,许多基于α稳定分布信号处理算法中参数的选取大多依赖α的先验知识或估计值。从计算复杂度层面分析比较了样本分位数法、对数法和负阶矩法3种常用的特征指数估计方法,并利用计算机仿真得出了3种估计方法对α值的估计结果和估计精度,给出了3种估计方法的使用条件,为α稳定分布的应用提供了便利。The a-stable distribution can be conveniently described by its characteristic func- tion, which is completely determined by four parameters. aracteristic exponent a is the most important parameter. Usually, parameters in many signal processing algorithms based on a-stable distribution mostly depend on the In this paper, the method of sam negative-order computatiOnal exponents are priori knowledge ple fractiles, the moments are discussed. The three or estimated value of characteristic exponent. logari ethod mic method and the method based on s are analyzed and complexity. The estimated results and estimation accuracies obtained by computer simulations. An advice is given for the compared for the ch in terms of aracterlstlcs conditions under which a particular estimation method should be used. This provides convenience in the application of a- stable distribution.

关 键 词：Α稳定分布 特征指数 样本分位数法 对数法 负阶矩法 

分 类 号：TN911.6[电子电信—通信与信息系统]