On generalized Schur and Levinson-Szego algorithms for quasistationary processes

We show how the well known reflection-coefficients parametrization of stationary second-order processes can be extended to quasi-stationary processes (i.e., whose covariance functions have finite displacement rank) by introducing sets of so-called generalized Schur parameters. Several properties of these generalized parameters will be studied, including covariance matching properties, generalized orthogonal polynomials, and the nontrivial question of when they can be used to specify whitening and modeling filters in lattice (or ladder) filter form.