Polynomial bounds for the linear prediction coefficients

 
see the original item page
in the repository's web site and access all digital files if the item*
share




1991 (EN)
Polynomial bounds for the linear prediction coefficients (EN)

Koukoutsis, E (EN)
Carayannis, G (EN)
Papaodysseus, C (EN)

N/A (EN)

The purpose of this paper is to introduce a new methodology for the study of the numerical behaviour of the Toeplitz system and the quantities involved in the linear prediction problem. For this reason, first, it is proved that the positive definiteness of the system matrix is equivalent to a set of constraints on the autocorrelation innovation, for which new explicit recursive formulae are given. Next, through these formulae, the minimum bounds of the absolute values of the linear prediction coefficients are computed, which are of the order of P(p/2). However, it is proved that, by imposing proper restrictions on the autocorrelation values, a linear or a polynomial bound of a desired order for the LP coefficients can be obtained. Finally, using the previous analysis, the ill-conditioning of the Toeplitz system and the sensitivity of the determinants of the corresponding matrices are discussed. (EN)

journalArticle

forward linear prediction, Toeplitz matrices (EN)
ill-conditioning (EN)
linear predictors (EN)
Signal Processing--Analysis (EN)
numerical behaviour (EN)
PARCORs (EN)
positive definiteness (EN)
Polynomials bounds (EN)
Linear prediction (EN)
Signal Filtering and Prediction (EN)
Mathematical Techniques (EN)
autocorrelation coefficients (EN)
Autocorrelation (EN)

Εθνικό Μετσόβιο Πολυτεχνείο (EL)
National Technical University of Athens (EN)

Signal Processing (EN)

1991


ELSEVIER SCIENCE BV (EN)



*Institutions are responsible for keeping their URLs functional (digital file, item page in repository site)