Recognition of sine wave modeled consonants by normal hearing and hearing -impaired individuals.
Item
-
Title
-
Recognition of sine wave modeled consonants by normal hearing and hearing -impaired individuals.
-
Identifier
-
AAI3144077
-
identifier
-
3144077
-
Creator
-
Balachandran, Rupa.
-
Contributor
-
Adviser: Arlene Neuman
-
Date
-
2004
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Health Sciences, Audiology | Psychology, Experimental | Physics, Acoustics
-
Abstract
-
Sine wave modeling is a parametric tool for representing the speech signal with a limited number of sine waves. It involves replacing the peaks of the speech spectrum with sine waves and discarding the rest of the lower amplitude components during synthesis. It has the potential to be used as a speech enhancement technique for hearing-impaired adults. The present study answers the following basic questions: (1) Are sine wave synthesized speech tokens more intelligible than natural speech tokens? (2) What is the effect of varying the number of sine waves on consonant recognition in quiet? (3) What is the effect of varying the number of sine waves on consonant recognition in noise? (4) How does sine wave modeling affect the transmission of speech feature in quiet and in noise? (5) Are there differences in recognition performance between normal hearing and hearing-impaired listeners?;VCV syllables representing 20 consonants (/p/, /t/, /k/, /b/, /d/, /g/, /f/, /theta/, /s/, /∫/, /v/, /z/, /t∫/, /dy/, /j/, /w/, /r/, /l/, /m/, /n/) in three vowel contexts (/a/, /i/, /u/) were modeled with 4, 8, 12, and 16 sine waves. A consonant recognition task was performed in quiet, and in background noise (+10 dB and 0 dB SNR). Twenty hearing-impaired listeners and six normal hearing listeners were tested under headphones at their most comfortable listening level. The main findings were: (1) Recognition of unprocessed speech was better that of sine wave modeled speech. (2) Asymptotic performance was reached with 8 sine waves in quiet for both normal hearing and hearing-impaired listeners. (3) Consonant recognition performance in noise improved with increasing number of sine waves. (4) As the number of sine waves was decreased, place information was lost first, followed by manner, and finally voicing. (5) Hearing-impaired listeners made more errors then normal hearing listeners, but there were no differences in the error patterns made by both groups.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
