5pSC11. Modeling the categorization of a large F1--F2--F3 continuum by Finnish listeners.

Session: Friday Afternoon, May 17

Time: 3:30

Author: Terrance M. Nearey
Author: Michael Kiefte
Location: Dept. of Linguist., Univ. of Alberta, Edmonton, AB T6G 0A2, Canada


A continuum of 972 vowels was synthesized, each of which was 115 ms in duration with a falling F0 contour (125--100 Hz). F1 ranged (in 0.5- Bark steps) from 250 to 760 Hz, F2 from 750 to 2260 Hz and F3 from 1360 to 3080 Hz. F4 and F5 were fixed at 3500 and 4500 Hz, respectively. (Constraints were placed on formant separations to ensure relatively natural stimuli.) This continuum was previously categorized by a group of English speakers [T. M. Nearey and M. Kiefte, J. Acoust. Soc. Am. 96, 3284(A) (1994)]. The same stimuli were presented to 14 speakers of Finnish at the University of Turku. Subjects categorized these stimuli with labels for the nine Finnish vowels, / i, e, y, o, (ae ligature), u, o/. Results of logistic regression analysis will be presented and compared with previous findings for English. [Work supported by SSHRC. The assistance of Ulla Fisher and Dr. Olli Aaltonen of the University of Turku in collecting the Finnish data is gratefully acknowledged.]

from ASA 131st Meeting, Indianapolis, May 1996