ARMA digital lattice filter based on new criterion

In this paper, a new ARMA digital lattice filter is proposed. First, a new criterion is defined for designing the ARMA digital lattice filter from given stochastic data. Based on this criterion, an ARMA parameter estimation algorithm is developed. This algorithm can estimate the ARMA parameters with little calculation cost. Then, from the fast algorithm of the ARMA parameter estimation, two elementary sections of the ARMA digital lattice filter are invented. Any ARMA model with an arbitrary AR order and an arbitrary MA order can be realized by using these elementary sections. In this paper, the proposed ARMA lattice filter is compared with other fast algorithms of ARMA parameter estimation and other ARMA lattice filters. Experimental results obtained from applying this filter to model reduction problems are shown. From these results, it is shown that this ARMA lattice filter is quite useful for ARMA digital signal processing.