A time‐varying parameter estimation of a minimum realization ARMA lattice model

Abstract

AbstractThis paper proposes a time‐varying parameter estimation of a minimum realization ARMA lattice model. The model consists of four types of elementary sections. Since each elementary section has only a design parameter after normalization, its characteristics can be represented more compactly. To estimate time‐varying characteristics, several methods with the constant weighting factor have been proposed.This paper shows the limits of the methods using the constant weighting factor and proposes an estimation method with a new variable weighting factor. The new method directly weights observed signals and can arbitrarily change the weighting factor at any time.