A test of nonlinear autoregressive models

Author

Mao, Shi-Yi ; Lin, Pin-Xing

Author_Institution

Dept. of Electron. Eng., Beijing Inst. of Aeronaut. & Astronaut., China

fYear

1988

fDate

11-14 Apr 1988

Firstpage

2276

Abstract

A study on testing the appropriateness of a particular structure selection and design for block-oriented nonlinear models is presented. Block-oriented nonlinear models characterize some features of Volterra kernels and extract only particular higher-order statistical information. Correlation between error and all possible products of data can be used to determine which kind of block-oriented nonlinear model is appropriate. Different structures are concerned with different higher-order statistics. The prediction error performance would be improved only of a correct model is chosen. The results of simulation studies are included to illustrate the validity of the conclusions

Keywords

filtering and prediction theory; parameter estimation; signal processing; statistical analysis; Hammerstein model; Volterra kernels; Wiener model; block-oriented nonlinear models; higher-order statistical information; nonlinear autoregressive models; parameter estimation; prediction error performance; signal modelling; Data mining; Design engineering; Electronic equipment testing; Higher order statistics; Kernel; Nonlinear systems; Predictive models; Signal analysis; System identification; Time series analysis;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1988. ICASSP-88., 1988 International Conference on

Conference_Location

New York, NY

ISSN

1520-6149

Type

conf

DOI

10.1109/ICASSP.1988.197091

Filename

197091