Optimum Design of Frequency-Response-Masking Filters Using Convex-Concave Procedure

Abstract

The class of frequency-response-masking (FRM) filters, first investigated by Y.C. Lim in 1980s, has shown to provide considerably improved performance for the design of finite impulse response (FIR) filters with narrow transition bandwidth relative to its conventional counterparts. In this project report, we present a design method for FRM filters by using a technique known as convex-concave procedure (CCP). For illustration purposes, we begin by reviewing several basic concepts and properties of FIR filters as well as a popular design technique based on windowed Fourier series. This is followed by a chapter to introduce the structure of basic FRM filters and to explain how they work. The CCP is then studied in the next chapter as an effective heuristic for nonconvex constrained optimization problems. In the last chapter, we present a formulation procedure that converts a frequency-weighted minimax design of FRM filters to a problem which can be solved by iteratively performing CCP. Simulation runs are included to illustrate the proposed algorithm and evaluate the design performance in comparison with those achieved by several existing methods.