ARMA digital lattice filters based on geometric approach

抄録

This paper discusses various ARMA digital lattice filters. The design procedures of these filters have been proposed as various forms. However, if we notice the orthogonal method based on a linear estimation theory, all design procedures are treated with an understandable coherent form. First, this paper examines the ARMA lattice filter whose section increases either AR order or MA order by one independently. The filter section is regarded as elementary. In order to derive this section, the subspace used in the or-thogonal projection method is carefully processed with the other subspaces. In other words, this paper shows the implicit dependency in each elementary section. In addition, this paper develops the ARMA filter associated with the Mullis-Roberts' algorithm. This algorithm estimates the AR parameters for the ARMA reference model. Thus, it used to have nothing to do with the construction of the lattice filter. However, if an ap-propriate subspace is selected, we can design a Mullis-Roberts' algorithm type ARMA lattice filter. In this paper, all relations in the design procedures become clearto under-stand them easily.