Nonlinear FIR Filter Design by Genetic Algorithm


In this paper, an approach to digital filter design which reduces the number of design steps is proposed, whereby the multi-objective optimisation ability of the Genetic Algorithm is utilised to simplify the design procedure. The technique is designed to produce a range of solutions with different performance tradeoffs, from which the designer can select the most appropriate. An analysis of the parameter space for an FIR filter design problem is presented, with a number of investigations into its difficulty. Reasons for the failure of the GA to produce satisfactory results are discussed.