Applying factorization to increase the resolving ability of the parametric estimation of the power spectral density

Abstract

We consider a possibility of the factorization of parametric spectral power density (PSM) estimation of a random process based on autoregressive linear prediction model to increase the spectrum resolution. Factorization refers to the decomposition of the multimode PSM into simpler single-mode components. Factorization makes it possible not only to decompose a complex multimode PSM into simple single-mode components, but also to analyze more accurately the low-, medium- and high-frequency components of the SPM of a random process. The main attention is paid to the study of the problem of increasing the resolving power of SPM estimation by its factorization by the Yule-Walker and Berg method.