A Capability Improvement for Design of CSD Coefficient FIR Filters Using Parameter-free GA

Abstract

In this paper, a design method of CSD (Canonic Signed Digit) coefficient FIR (Finite Impulse Response) filters using parameter-free genetic algorithm (PfGA) is proposed. In GA (Genetic Algorithm), we have a large amounts of difficulties for settings many parameters involved in the GA before the design. Such a task often forces excessive effort to the filter designers. In addiion, the general GA requires many individuals for increasing the diversification ability of search. This fact causes a large amounts of computational time for design. In order to overcome those difficulties, PfGA was proposed, in which any GA parameters are not needed to set in advance. As a result, the task for the parameter verification is not required. However, a detail effectiveness of this strategy for CSD coefficient FIR filter design problem which has a complex structure is not known. In this paper, several design examples are shown to present the practical effectiveness. The outcomes of this paper may give a good opportunity to free a lots of difficulties from the filter designers.