Stable multi-dimensional model reduction and IIR filter design

Abstract

It is well-known that model reduction techniques applied to stable multi-dimensional Linear Shift-Invariant (LSI) systems with Infinite-extent Impulse Response (IIR) do not necessarily guarantee a stable reduced system. Several conditions exist to check stability a posteriori. In this paper we outline a new technique that guarantees, a priori, that the system or filter is stable. In Section 1 we establish the necessary notation and definitions to deal with multi-dimensional systems and filters. Section 2 introduces the technique of multivariate Pade-type approximation and deals with stability. In Section 3 we illustrate the use of the newly proposed technique in filter design.