This project explores the perceptual consequences of articulatory differences between """"""""clear"""""""" and """"""""conversational"""""""" speech, in particular, spectral change effects reflected by, e.g., formant slopes into and out of consonants possibly stemming from greater articulatory """"""""effort"""""""", in the clear speech case. The focus on these effects will separate them from duration effects and determine their relative contribution to the greater intelligibility of clear speech. These studies are implemented by a signal processing algorithm that automatically alters the speech signal in ways resembling those employed by talkers when asked to speak clearly, without altering duration effects. Thus, applying the algorithm to conversational speech results in speech signal displaying exaggerated change, e.g., formants approach more extreme values, with resulting steeper formant transitions. These algorithm changes are produced by operating directly on speech parameters, without requiring segment recognition of any kind. Preliminary algorithm experiments found improved stop consonant identification in three-syllable stimuli masked by noise.
The specific aims of proposed research are to assess the potential of this signal processing algorithm (a) as a research tool and (b) as a form of speech enhancement for both normal and hearing-impaired listeners. The first stage in the project will establish a database of test consonants in four-syllable speech stimuli spoken in clear, conversational and also """"""""clear but rapid"""""""" modes. Various acoustic measures will determine the most effective algorithm parameters for transforming a signal spoken in conversational mode to one spoken in clear mode. Then the algorithm will be tested by experiments, using both normal-hearing and hearing-impaired listeners as subjects. The goal of this stage is to identify the spectral changes that are optimally effective for intelligibility across the consonant classes, for both subject populations. Further stages will study utterances of extending length and complexity. The goal of final stages is to generalize results to situations that more closely approximates the talker's and listener's task in normal communication.

Agency
National Institute of Health (NIH)
Institute
National Institute of Neurological Disorders and Stroke (NINDS)
Type
Research Project (R01)
Project #
1R01NS023249-01A1
Application #
3406521
Study Section
Communication Sciences and Disorders (CMS)
Project Start
1986-08-01
Project End
1989-07-31
Budget Start
1986-08-01
Budget End
1987-07-31
Support Year
1
Fiscal Year
1986
Total Cost
Indirect Cost
Name
University of Maryland College Park
Department
Type
Schools of Arts and Sciences
DUNS #
City
College Park
State
MD
Country
United States
Zip Code
20742