Abstract
An efficient algorithm for design and implementation of uniform and nonuniform digital filter banks is presented. By means of a single design parameter, a continuum of frequency spacings can be obtained. As an example, a frequency spacing similar to the critical bands of the human auditory system is possible, showing no significant deviations from critical band measurements. No numerical optimization technique to achieve a flat overall filter bank response is required, because this filter bank feature is inherent to the design method. The filter bank design is based on analog Lerner filters which exhibit sharp cutoff regions and are nearly linear phase. It is shown that a similar approach is feasible in the digital domain, which retains the neat properties of analog Lerner filters. In addition, filter banks can be efficiently implemented because filter sections can be shared between adjacent bandpass filters.

This publication has 6 references indexed in Scilit: