A new local search method to design finite precision FIR filters

Abstract

This paper proposes a new local search method for the design of FIR filters with low complexity finite precision coefficients based on the optimality criterion of peak constrained least squares. In order to improve search efficient and reduce invalid search steps, the proposed method adjusts the coefficient offsetting strategy conditionally and utilizes the geometric properties of the object function ) (h H that describes a hyper-ellipse. If the search point is inside ) (h H , it is selected, so the search trapped in an infeasible region is avoided. The tradeoff between coefficient complexity and PSR is got. Numerical experiments show that the proposed method holds better performance than the existed methods in terms of the search efficiency.