Hurst指数估计法中的修正方法研究  

Research of the correction methods in Hurst estimations

作　　者：朱灵蕾 姚远程[1] 姜军[2] 秦明伟[1] 

机构地区：[1]西南科技大学信息工程学院特殊环境机器人技术四川省重点实验室,四川绵阳621010 [2]西藏大学工学院,西藏拉萨850000

出　　处：《电子技术应用》2016年第7期103-106,110,共5页Application of Electronic Technique

基　　金：国防基础科研计划资助项目(B3120133002);西南科技大学创新团队基金(tdtk02);西南科技大学研究生创新基金资助项目(15ycx117)

摘　　要：Hurst参数是表征网络业务量自相似性的重要参数,对突发业务的Hurst参数进行准确的估计能直接影响网络流量控制和缓冲资源分配。详细给出4种Hurst参数估计方法的实现过程,并针对这4种估计方法进行修正。通过估计不同Hurst参数的自相似业务量来对比修正前后估计方法的精度,结果表明采用修正的方法各个估计方法的相对误差都有降低,其中R/S法降低了2个百分点,聚类方差法降低了8个百分点,周期图法和小波分析法相对误差降低了一个数量级。Hurst is an important parameter which can characterize the self similarity of network traffic,accurate estimating the Hurst parameters of the burst traffic can directly affects the network flow control and buffer resource allocation.This paper gives a detail implementation of the four Hurst parameter estimation methods,and proposes modified methods for these four estimation methods.The flour estimations before they are modified and after modified are compared though estimating the self similar traffic on different Hurst parameters,results shows that the relative error of each estimation methods reduced by using the modified method.Among them,the R / S method is reduced by 2 percentage points,and the aggregated variance method decreases by 8 percentage points,the relative error of the periodogram method and the wave estimator is reduced by an order of magnitude.

关 键 词：自相似性 分形高斯噪声 赫斯特参数 R/S法 聚类方差 小波 

分 类 号：TN915[电子电信—通信与信息系统] TP393[电子电信—信息与通信工程]