Response of Analyzers of the Bank‐of‐Filters Type to Signals Generated by Vowel Sounds

A number of different kinds of filters have been used, or proposed for use, in speech analyzers of the bank‐of‐filters type. These include single‐tuned, cascaded single‐tuned, double‐tuned, cascaded double‐tuned, Butterworth, Gaussian, and Tchebycheff. Consider a group of such analyzers having identical filter bandwidth and center‐frequency characteristics; each analyzer in the group is comprised of filters of a different type. This paper presents a comparison of the analyzers in this group on the basis of errors in location of formants of vowel sounds. In this investigation, the bank‐of‐filters, as well as the input signals to the analyzers, was synthesized on a digital computer.