Design of 1-D and 2-D IIR digital filters using IRLS technique.

Abstract

In this thesis a design technique called iteratively reweighted least-squares (IRLS) is used to design 1-D and 2-D IIR digital filters in the Lp norm sense. In this technique by selecting the weights in each iteration based on the error in previous iterations, a solution to the Lp approximation problem is found by solving a weighted least-squares problem. In order to successfully design the IIR digital filters using the IRLS technique a convergence control parameter is determined in a manner suitable for the IIR filter design problem. A comparison of various 1-D and 2-D IIR filters designed using the IRLS technique and the Davidon Fletcher Powell (DFP) unconstrained optimization technique is also made, and the relative computational complexity of the two techniques is discussed. Comparisons with some other recent techniques both for 1-D and 2-D IIR digital filters are also presented.