AR model parameter estimation: from factor graphs to algorithms

Author

Korl, Sascha ; Loeliger, Hans-Andrea ; Lindgren, Allen G.

Author_Institution

Signal and Information Processing Laboratory, ETH Zurich, Switzerland

Volume

5

fYear

2004

fDate

17-21 May 2004

Abstract

The classic problem of estimating the parameters of an auto-regressive (AR) model is considered from a graphical model viewpoint. A number of practical parameter estimation algorithms - some of them well known, others apparently new - are derived as "summary propagation" in a factor graph. In particular, we demonstrate the joint estimation of AR coefficients, innovation variance, and noise variance.

Keywords

Kalman filters; Monte Carlo methods; autoregressive processes; graph theory; least mean squares methods; parameter estimation; AR coefficients; AR model parameter estimation; Gaussian AR models; Kalman filters; LMS-type algorithms; autoregressive models; factor graph summary propagation; factor graphs; graphical models; innovation variance; message passing algorithms; noise variance; particle filters; Error correction codes; Gaussian noise; Graphical models; Information processing; Laboratories; Parameter estimation; Signal processing; Signal processing algorithms; State-space methods; Technological innovation;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 2004. Proceedings. (ICASSP '04). IEEE International Conference on

ISSN

1520-6149

Print_ISBN

0-7803-8484-9

Type

conf

DOI

10.1109/ICASSP.2004.1327159

Filename

1327159