A novel technique for DCGA optimization of guaranteed BIBO stable IIR-based FRM digital filters over the CSD multiplier coefficient space

This paper presents a novel diversity-controlled (DC) genetic algorithm (GA) for the design and rapid optimization of frequency-response masking (FRM) digital filters over the CSD multiplier coefficient space. The resulting FRM digital filters incorporate bilinear-LDI IIR interpolation subfilters realized as a parallel combination of a pair of allpass digital networks. A novel LUT scheme is developed to ensure that the FRM digital filters under consideration are automatically BIBO stable throughout the course of DCGA optimization. The salient feature of the proposed LUT scheme is that it makes no recourse to slack variables for referencing the values of the CSD multiplier coefficients. The DCGA optimization fitness function includes not only the magnitude but also the group-delay frequency-response of FRM digital filters so as to minimize phase distortion caused by the IIR interpolation subfilters. An example is given to illustrate the application of the proposed DCGA optimization to the design of a lowpass FRM digital filter incorporating a seventh-order bilinear-LDI interpolation subfilter.