A neural network approach to FIR filter design using frequency-response masking technique

This paper presents a new optimization method for the design of various frequency-response-masking (FRM)-based linear-phase finite-impulse response (FIR) digital filters. The method is based on a batch back-propagation neural network algorithm (NNA), which is taken as a variable learning rate mode. In order to reduce the complexity, the following two-step optimization technique is proposed. At the first step, an initial FRM filter is designed by alternately optimizing the sub-filters. This solution is then used as a start-up solution for further optimization. At the second step, the coefficients of overall sub-filters are optimized simultaneously by the NNA. Algorithm details for the design of basic and multistage FRM filters are presented to show that the proposed approach offers a unified design framework for a variety of FRM filters. Some examples taken from the literatures are included and the results show that the proposed algorithm can design better FRM filters than several existing methods.