The Use of Automatic Speech Recognition as a Speech Clarifier
Lead Institution: Duke University
|
|
| PURPOSE |
| Our goal is to perform the research necessary to engineer a
portable system with text and/or speech output to clarify dysarthric
speech. |
| TARGET POPULATION |
| Individuals with mild, moderate, and severe speech disorders
who are reliably understood by familiar communication partners,
but who have difficulty communicating in noisy environments or
with unfamiliar communication partners |
| PROBLEM |
|
Automatic speech recognition (ASR) systems, such as DragonDictate®,
VoiceExpress, and ViaVoice™ are being used with increasing
frequency by the general population and individuals with physical
disabilities. While there has been significant improvement in
the area of ASR, these commercially-available systems do not
work well for individuals with imperfect speech.
|
| PROGRESS |
|
Data collection began in October 2001 with several sample sets
of simulated user data collected from research staff. Speech
samples need to be collected in a specific way in order to be
used to build models. Conditions including room noise, subject
position, and microphone location all affect the quality of
the sample. Staff developed data collection protocols, and pilot
data was collected in February and March of 2002. Sample data
sets will be sent to Navy Air Warfare Center (NAWC) in May 2002
for models to be built.
|
| KEY FINDINGS |
|
None to date
|
| PARTNERS |
|
(the links below will open in a new window)
|
| FOR MORE INFORMATION |
Kevin M. Caves, BSME : kevin.caves@duke.edu
Frank DeRuyter, PhD : fdr-rerc@mc.duke.edu
|
|
|
