Representation of The First Formant In Speech Recognition And In Models of The Auditory Periphery

Authors

  • Dennis H. Klatt

Abstract

The frequency and amplitude of' the first formant are not easy to measure as fundamental frequency (f0) varies in speech. Perceptual data indicate that the auditory system is not bothered by changes to f0, but processing strategies used in speech recognition, such as linear prediction, filterbank analysis, and the synchrony spectrum are seriously perturbed as f0 varies. The irrelevant variation makes it difficult/unreliable to perform phonetic comparisons between similar vowels based on simple ideas of pattern similarity. Of the possible solutions to this problem considered here, the one of greatest practical attraction is to implement a synchrony spectrum representation of vowel-like speech sounds, and a "learned pattern equivalence" approach to vowel phonetic-quality equivalence across different fundamental frequencies.

Downloads

Published

1986-07-21

How to Cite

1.
Klatt DH. Representation of The First Formant In Speech Recognition And In Models of The Auditory Periphery. Canadian Acoustics [Internet]. 1986 Jul. 21 [cited 2023 Feb. 8];14(3 bis):5-7. Available from: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/3495

Issue

Section

Proceedings of the Acoustics Week in Canada