PARAMETRIC ESTIMATION METHOD OF THE POWER SPECTRAL DENSITY BASED ON MODELING OF THE LAPLACE REPRESENTATION OF THE AUTOCORRELATION FUNCTION

摘要

In this paper we propose a new parametric method for estimation of a power spectral density. The essence of this method consists of construction of the structured model for the Laplace representation of an autocorrelation function of an investigated random process. In particular we used as a model the fractional rational function which has polynomials in complex frequency with the real coefficients in numerator and denominator. Calculation of best values of coefficients of this model is provided due to solution of the overdetermined initial system of equations. On the basis of this solution we have developed digital recurrent algorithms for estimation of coefficients of the model of the Laplace representation of an autocorrelation function. The order of model is selected on the basis of the analysis of values of the error sum of squares. Results of computer simulation for estimation of a power spectral density of the additive mixture of sinusoidal components are presented.