ARMA MODELING USING CUMULANT AND AUTOCORRELATION STATISTICS.

G. B. Giannakis, J. M. Mendel, W. Wang

Research output: Contribution to journalConference articlepeer-review

15 Scopus citations

Abstract

One-dimensional cumulant and autocorrelation output statistics are combined to form an overdetermined system of equations whose least-squares solution yields the coefficients of an autoregression moving average (ARMA) model. The driving input noise is assumed to be non-Gaussian and white. The ARMA model is allowed to be nonminimum-phase and even to contain all-pass factors. The special cases of AR and MA models are also included. The overdetermined nature of the method makes the solution practical for moderate output data lengths, when additive white Gaussian noise is considered. Simulations illustrate that the approach performs very well even at low signal-to-noise ratios.

Original languageEnglish (US)
Pages (from-to)61-64
Number of pages4
JournalICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
StatePublished - Jan 1 1987

Fingerprint Dive into the research topics of 'ARMA MODELING USING CUMULANT AND AUTOCORRELATION STATISTICS.'. Together they form a unique fingerprint.

Cite this