Formant transitions as partly distinctive invariant properties in the identification of voiced stops

Authors

  • T.M. Nearey Dept. of Linguistics, Alberta Univ., Edmonton, Alta., Canada
  • S.E. Shammass Dept. of Linguistics, Alberta Univ., Edmonton, Alta., Canada

Keywords:

speech analysis and processing, formant transitions, speech analysis, partly distinctive invariant properties, voiced stops, F2 trajectories, Canadian English, regression line, consonant, category regression lines, statistical classification

Abstract

The F2 trajectories for /b, d, g/ in /CVd/ syllables are often summarized by the initial F2 frequency (F2i) and that of the `steady-state vowel' (F2v). Trajectories were measured for 660 Canadian English/CVd/syllables (3 stops×11 vowels×10 speakers×2 repetitions). Plots for each stop (vowels pooled) indicated a strong linear relationship between F2i and F2v. A regression line fitted to each plot represents an invariant relational property of the corresponding consonant. F2 trajectories are not sufficient to specify the stops uniquely, since the lines for the three consonants intersect (indicating category overlap). However, the slopes and intercepts for the three consonants are distinct and thus represent partly distinctive invariant properties or partial invariants. Similar patterns obtain for F3. Use of partial invariants of F2/F3 trajectories in a classification algorithm (based on minimum distance from category regression lines) result in an identification rate of over 70%, which compares favorably with a number of other statistical classification schemes. Possible extensions of this approach and relationships to aspects of perception are discussed

Additional Files

Published

1987-12-01

How to Cite

1.
Nearey T, Shammass S. Formant transitions as partly distinctive invariant properties in the identification of voiced stops. Canadian Acoustics [Internet]. 1987 Dec. 1 [cited 2024 Jun. 20];15(4):17-24. Available from: https://jcaa.caa-aca.ca/index.php/jcaa/article/view/581

Issue

Section

Technical Articles