Polynomial expansion of the probability density function about gaussian mixtures

Author

Cavalcante, C.C. ; Mota, J.C.M. ; Romano, J.M.T.

Author_Institution

GTEL/DETI/CT/UFC, C.P. 6005, Campus do Pici, CEP: 60.455-760, Fortaleza-CE, Brazil

fYear

2004

fDate

Sept. 29 2004-Oct. 1 2004

Firstpage

163

Lastpage

172

Abstract

A polynomial expansion to probability density function (pdf) approximation about Gaussian mixture densities is proposed in this paper. Using known polynomial series expansions we apply the Parzen estimator to derive an orthonormal basis that is able to represent the characteristics of probability distributions that are not concentrated in the vicinity of the mean point such as the Gaussian pdf. The blind source separation problem is used to illustrate the applicability of the proposal in practical analysis of the dynamics of the recovered data pdf estimation. Simulations are carried out to illustrate the analysis

Keywords

Gaussian distribution; blind source separation; function approximation; polynomial approximation; Gaussian mixture densities; Parzen estimator; blind source separation problem; orthonormal basis; polynomial series expansion; probability density function approximation; probability distributions; Analytical models; Blind source separation; Interference; Polynomials; Postal services; Probability density function; Probability distribution; Proposals; Signal processing; Source separation;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning for Signal Processing, 2004. Proceedings of the 2004 14th IEEE Signal Processing Society Workshop

Conference_Location

Sao Luis

ISSN

1551-2541

Print_ISBN

0-7803-8608-4

Type

conf

DOI

10.1109/MLSP.2004.1422970

Filename

1422970