Lattice algorithms for recursive ARMAX identification

Abstract

Lattice algorithms are fast recursive identification algorithms, that estimate models of several orders simultaneously. Up to now lattice algorithms were only available for AR, ARX and ARMA models. In this paper lattice algorithms for ARMAX models are proposed. Emphasis is put on the practical advantages of these algorithms over conventional (non-lattice) recursive identification methods. The number of computations is much less, and the convergence in the presence of noise is drastically improved.