Criterion of Significance Level for Selection of Order of Spectral Estimation of Entropy Maximum
- 作者: Savchenko V.1, Savchenko A.2
-
隶属关系:
- Nizhny Novgorod State Linguistic University
- National Research University Higher School of Economics
- 期: 卷 62, 编号 5 (2019)
- 页面: 223-231
- 栏目: Article
- URL: https://journals.rcsi.science/0735-2727/article/view/177312
- DOI: https://doi.org/10.3103/S0735272719050042
- ID: 177312
如何引用文章
详细
It is researched a wide class of parametric estimations of power spectral density based on principle of entropy maximum and autoregression observation model. At that there is distinguished the key parameter which is used model order. It is considered a problem of a priori uncertainty when true value of order is a priori unknown. It is proposed a new criterion for definition of order using finite sampling volume with purpose of overcome of the drawbacks of existing algorithms in conditions of small sampling. The principle of guaranteed significance level in a problem of complex statistic hypothesis verification is a basic principle of this criterion. In contrast to criteria of AIC, BIC, etc. this criterion is not related to determination of measurements inaccuracy, since it uses a conception of “significance level” of formed solution only. The efficiency of proposed criterion is researched theoretically and experimentally. An example of its application in a problem of spectral analysis of voice signals is considered. Recommendations about its practical application in the systems of digital signal processing are given.
作者简介
V. Savchenko
Nizhny Novgorod State Linguistic University
编辑信件的主要联系方式.
Email: vvsavchenko@yandex.ru
俄罗斯联邦, Nizhny Novgorod
A. Savchenko
National Research University Higher School of Economics
编辑信件的主要联系方式.
Email: avsavchenko@hse.ru
俄罗斯联邦, Nizhny Novgorod