Multicriteria design of digital filter with evolutionary optimization

A multicriteria design technique for linear phase finite impulse response filter, based on multiobjective evolutionary optimization, is presented. In the proposed method, filter coefficients are optimized during the simulated evolution, which is driven to satisfy the given requirements in a frequency domain, as well as application specific constraints on a time domain signal. Multiobjective evolutionary algorithm combines concepts of non- dominated sorting with a selection based on the hypervolume measure. The algorithm efficiency is evaluated for the design of channel filter for orthogonal frequency division multiplexing system, where the error vector magnitude of the output signal is incorporated into the filter design process. The obtained results show that the idea of filter optimization with combined frequency and time domain requirements can be a valuable extension to the available filter design methods.