Effects of inter-talker differences on speech perception.
Item
-
Title
-
Effects of inter-talker differences on speech perception.
-
Identifier
-
AAI9924831
-
identifier
-
9924831
-
Creator
-
Mulhearn, Bethany Ann.
-
Contributor
-
Adviser: Arthur Boothroyd
-
Date
-
1999
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Health Sciences, Audiology | Psychology, Cognitive
-
Abstract
-
Clinical measures of the speech perception of hearing-impaired listeners typically use one talker who may not be representative of the "average talker". While inter-talker differences in the acoustic properties of speech are known, the contribution of these differences to the confidence limits of speech perception measures has not been established. The goals of this study were: (a) to measure inter-talker differences in terms of key parameters of the distribution of information across frequency, (b) to establish the acoustic correlates of these differences, and (c) to estimate their potential contribution to the confidence limits of test measures.;Consonant-vowel-consonant words were produced by three adult males and three adult females. Phoneme recognition was measured in normally hearing adults as a function of cut-off frequency for low- and high-pass filtering. For each talker, the following was derived: (a) "cross-over frequency" for the two functions, i.e., the point where the frequency spectrum is divided into two equally intelligible parts, (b) "half-band phoneme intelligibility", i.e., phoneme recognition at cross-over, and (c) "full-band phoneme intelligibility", derived from half-band intelligibility by application of the articulation index theory.;Mean cross-over frequency for male talkers was 1599 Hz, with 95% confidence limits of +/--117 Hz. Mean cross-over frequency for female talkers was 1806 Hz, with 95% confidence limits of +/--115 Hz. The difference of 13% was statistically significant (p = 0.027). There were also inter-subject differences of half-band and full-band intelligibility, with no evidence that these differences were gender-related. The acoustic data were consistent with the hypothesis that higher cross-over frequency for females results from higher formant frequencies. Using data obtained under low-pass filtering, it was found that inter-talker differences contribute significantly to the confidence limits of phoneme recognition. That contribution is considerably less, however, than that of inherent test-retest variability of speech recognition scores when they are based on the small number of test items typical of clinical practice. If inherent variability can be reduced, it may then be useful to seek ways of minimizing the effects of inter-talker differences.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
