LINGUIST List 15.16

Mon Jan 12 2004

Diss: Phonetics: Adank: 'Vowel Normalization...'

Editor for this issue: Takako Matsui <takolinguistlist.org>


Directory

  1. p.adank, Vowel Normalization: a perceptual-acoustic study of Dutch vowels

Message 1: Vowel Normalization: a perceptual-acoustic study of Dutch vowels

Date: Mon, 5 Jan 2004 07:43:34 -0500 (EST)
From: p.adank <p.adanklet.kun.nl>
Subject: Vowel Normalization: a perceptual-acoustic study of Dutch vowels

Institution: University of Nijmegen
Program: Department of linguistics
Dissertation Status: Completed
Degree Date: 2003

Author: Patti Adank 

Dissertation Title: Vowel Normalization: a perceptual-acoustic study
of Dutch vowels

Linguistic Field: Phonetics 
Subject Language: Dutch (code: DUT )

Dissertation Director 1: Roeland van Hout
Dissertation Director 2: Roel Smits

Dissertation Abstract: 

In sociolinguistics, language variation in vowel sounds is typically
studied using phonetic transcription. Phonetic transcription is
carried out by expert listeners, who are capable of perceptually
separating (socio-) linguistic variation from anatomical/physiological
speaker-related characteristics. However, phonetic transcription is a
very laborious task and its reliability is questionable. In phonetics,
vowel normalization procedures were designed to separate linguistic
variation from anatomical/physiological variation in acoustic
measurements.

In this thesis, it is evaluated whether vowel normalization procedures
are suitable for use in sociolinguistics, for studying language
variation in vowels. It is described how 12 procedures for vowel
normalization were compared. This comparison consists of two
parts. First, the procedures were evaluated in the acoustic domain by
applying them to acoustic (formant and fundamental frequency)
measurements of a database of vowels pronounced by 160 Dutch teachers
from the Netherlands and Flanders. The procedures were evaluated on
how well they preserved (socio-) linguistic variation and minimized
anatomical/physiological variation. Second, the normalization
procedures were compared in a perceptual-acoustic comparison; a
comparison with perceptual judgments made by expert listeners.

The normalization procedures were applied to the acoustic measurements
of vowel tokens from 20 of the 160 Dutch teachers. These tokens were
also judged by expert listeners on their perceived (socio-) linguistic
characteristics, i.e., tongue height, tongue position, and lip
rounding in a listening experiment. Next, the transformed acoustic
measurements were compared to the perceptual judgments using linear
regression analysis, to evaluate how well the output of the
normalization procedures modeled the perceived (socio-) linguistic
characteristics of the vowel tokens.

These two comparisons of the 12 normalization procedures show that
procedures that include information (e.g., the mean or standard
deviation of the formant measurements) about other vowels produced by
a speaker are most successful in the acoustic evaluation as well as in
the perceptual-acoustic comparison.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue