Autoassociative neural networks and noise filtering

Author

Dorronsoro, José R. ; López, Vicente ; Cruz, Carlos Santa ; Sigüenza, Juan A.

Author_Institution

Dept. de Ingenieria Informatica, Univ. Autonoma de Madrid, Spain

Volume

51

Issue

5

fYear

2003

fDate

5/1/2003 12:00:00 AM

Firstpage

1431

Lastpage

1438

Abstract

We introduce linear autoassociative neural (AN) network filters for the removal of additive noise from one-dimensional (1-D) time series. The AN network will have a (2M+1)×L×(2M+1) architecture, and for M fixed, we show how to choose the optimal L value and output coordinate from square error estimates between the AN filter outputs and the clean series. The frequency response of AN filters are also studied, and they are shown to act as matched band filters. A noise variance estimate is also derived from this analysis. We numerically illustrate their behavior on two examples and also compare their theoretical performance with that of optimal Wiener filters.

Keywords

filtering theory; matched filters; neural nets; noise; time series; 1D time series; additive noise removal; autoassociative neural networks; clean series; eigenfilter structure; frequency response; linear autoassociative neural network filters; matched band filters; noise filtering; noise variance estimate; one-dimensional time series; optimal Wiener filters; output coordinate; square error estimates; time series analysis; time-invariant linear filters; variance estimation; Adaptive filters; Covariance matrix; Delay effects; Filter bank; Filtering; Matched filters; Neural networks; Nonlinear filters; White noise; Wiener filter;

fLanguage

English

Journal_Title

Signal Processing, IEEE Transactions on

Publisher

ieee

ISSN

1053-587X

Type

jour

DOI

10.1109/TSP.2003.810276

Filename

1194429