A Second-Order Mapping PCA Based Intrinsic Dimension Estimate for Nonlinear Time Series

Author

Huang, Xiaolin ; Ning, Xinbao

Author_Institution

Key Lab. of Modern Acoust., Nanjing Univ., Nanjing

Volume

1

fYear

2008

fDate

18-20 Oct. 2008

Firstpage

274

Lastpage

279

Abstract

The delay vectors of nonlinear time series are mapped into a second-order space and the intrinsic dimension of the underlying dynamics is estimated by the saturation of the nonvanishing eigenvalues of the covariant matrix as the embedding dimension increases. This method needs less data and has certain noise resistant degree.

Keywords

covariance matrices; delays; eigenvalues and eigenfunctions; nonlinear dynamical systems; nonlinear estimation; time series; covariant matrix; delay vector; intrinsic dimension estimation; noise resistant degree; nonlinear time series; nonvanishing eigenvalue; primary component analysis; second-order mapping; Algorithm design and analysis; Covariance matrix; Delay effects; Delay estimation; Eigenvalues and eigenfunctions; Matrix decomposition; Phase estimation; Predictive models; Principal component analysis; Time series analysis; ECG; Nonlinear time series; delay embedding; intrinsic dimension; saturating embedding dimension; second-order mapping PCA;

fLanguage

English

Publisher

ieee

Conference_Titel

Natural Computation, 2008. ICNC '08. Fourth International Conference on

Conference_Location

Jinan

Print_ISBN

978-0-7695-3304-9

Type

conf

DOI

10.1109/ICNC.2008.119

Filename

4666853