Evolutionary Algorithm versus Variable Metric Method in Digital FIR Filter Design

In this paper we analyze the filter design task and interprete it as an optimization problem. We verify the possibility of applying good local optimization method (quasi-Newton - QN) and a robust global optimization method (evolutionary algorithm - EA). According to the results of the computer simulation, we conclude that a hybrid solution that combines these two method yields filters of the amplitude characteristic that is better than yielded by the window based design method and better than generated by the QN method alone.