ARMA digital lattice filter based on a linear prediction theory

Abstract

AbstractThis paper proposes some elementary sections of an ARMA lattice filter. an ARMA lattice filter with an arbitrary AR order and an arbitrary MA order is presented. A conventional ARMA lattice filter has already been constructed by a method used to develop an AR lattice filter to a 2‐dimensional filter. Therefore the conventional filter is not suitable for a filter realization when an AR order and an MA order differ from each other, and for the design of a minimum realized ARMA model. the most important characteristic of lattice filters is an orthogonality between every prediction error calculated at each elementary section. Employing this characteristic, we employ the AR lattice filter in a speech analysis, synthesis, an equalizer, and so on. It is shown that the proposed ARMA lattice filter satisfies an orthogonality. Furthermore, the ARMA lattice inverse filter which is an analyzer for an observed waveform and the ARMA lattice filter which is a synthesizer for the waveform are presented.