Mutual information in the frequency domain
Mutual information in the frequency domain
复制标题
DOI：
10.1016/j.jspi.2006.06.026
复制
发表时间：
2007-03-01
影响因子：
0.9
通讯作者：
Guha, Apratim
中科院分区：
文献类型：
--
作者：
Brillinger, David R.;Guha, Apratim
互信息系数是经典相关系数的一个强有力的扩展。特别地，它们具有零的性质，当且仅当所涉及的分量在统计上彼此独立。这一特点可以证明是有用的准备工作，以建立模型。在这篇文章中，一个频域变形MI的发展和研究的二元平稳时间序列。作为一个科学的例子，周围的地震噪声数据集进行了研究，并推断缺乏独立的组件。M I对频率的依赖性的特征可以用来暗示统计依赖性的性质。(c)2006 Elsevier B.V.保留所有权利。
Coefficients of mutual information (MI) can provide powerful extensions of classical coefficients of correlation. In particular, they have the property of vanishing if and only if the components involved are statistically independent of each other. This characteristic can prove useful in preparatory work to model building. In this article a frequency domain variant of MI is developed and studied for bivariate stationary time series. As a scientific example an ambient seismic noise data set is studied and a lack of independence of the components inferred. The character of the dependence of the M I on frequency may be used to suggest the nature of the statistical dependence. (c) 2006 Elsevier B.V. All rights reserved.