Modeling coarticulation in continuous speech

Brian O. Bush, Alexander Kain

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Modeling coarticulation in speech has been largely limited to short sequences and/or limited phonetic context. We introduce a methodology for modeling both formant frequency and bandwidth in continuous speech, allowing examination of sentencelevel coarticulation. The model represents continuous trajectories as a combination of overlapping local trajectories, which are represented by a weighted-addition of acoustic event targets by sigmoidal coarticulation functions characterized by slope and position. Estimation is achieved using a combination of hill-climbing and grid-search, with global target, joint slope for identical contexts, and local position parameters. We evaluate model performance for two speakers using an intelligibility test that compares vocoded model output to a purely vocoded and a natural condition.

Original languageEnglish (US)
Title of host publicationINTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association
PublisherInternational Speech and Communication Association
Pages193-197
Number of pages5
StatePublished - 2014
Event15th Annual Conference of the International Speech Communication Association: Celebrating the Diversity of Spoken Languages, INTERSPEECH 2014 - Singapore, Singapore
Duration: Sep 14 2014Sep 18 2014

Other

Other15th Annual Conference of the International Speech Communication Association: Celebrating the Diversity of Spoken Languages, INTERSPEECH 2014
CountrySingapore
CitySingapore
Period9/14/149/18/14

Fingerprint

Slope
Trajectory
Hill Climbing
Target
Performance Model
Trajectories
Modeling
Overlapping
Acoustics
Speech analysis
Bandwidth
Grid
Methodology
Evaluate
Output
Model
Context
Speech
Coarticulation
Continuous Speech

Keywords

  • Coarticulation
  • Continuous speech
  • Formants

ASJC Scopus subject areas

  • Language and Linguistics
  • Human-Computer Interaction
  • Signal Processing
  • Software
  • Modeling and Simulation

Cite this

Bush, B. O., & Kain, A. (2014). Modeling coarticulation in continuous speech. In INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association (pp. 193-197). International Speech and Communication Association.

Modeling coarticulation in continuous speech. / Bush, Brian O.; Kain, Alexander.

INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association. International Speech and Communication Association, 2014. p. 193-197.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bush, BO & Kain, A 2014, Modeling coarticulation in continuous speech. in INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association. International Speech and Communication Association, pp. 193-197, 15th Annual Conference of the International Speech Communication Association: Celebrating the Diversity of Spoken Languages, INTERSPEECH 2014, Singapore, Singapore, 9/14/14.
Bush BO, Kain A. Modeling coarticulation in continuous speech. In INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association. International Speech and Communication Association. 2014. p. 193-197
Bush, Brian O. ; Kain, Alexander. / Modeling coarticulation in continuous speech. INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association. International Speech and Communication Association, 2014. pp. 193-197
@inproceedings{2014f54271e1466d8d2781f048db2801,
title = "Modeling coarticulation in continuous speech",
abstract = "Modeling coarticulation in speech has been largely limited to short sequences and/or limited phonetic context. We introduce a methodology for modeling both formant frequency and bandwidth in continuous speech, allowing examination of sentencelevel coarticulation. The model represents continuous trajectories as a combination of overlapping local trajectories, which are represented by a weighted-addition of acoustic event targets by sigmoidal coarticulation functions characterized by slope and position. Estimation is achieved using a combination of hill-climbing and grid-search, with global target, joint slope for identical contexts, and local position parameters. We evaluate model performance for two speakers using an intelligibility test that compares vocoded model output to a purely vocoded and a natural condition.",
keywords = "Coarticulation, Continuous speech, Formants",
author = "Bush, {Brian O.} and Alexander Kain",
year = "2014",
language = "English (US)",
pages = "193--197",
booktitle = "INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association",
publisher = "International Speech and Communication Association",

}

TY - GEN

T1 - Modeling coarticulation in continuous speech

AU - Bush, Brian O.

AU - Kain, Alexander

PY - 2014

Y1 - 2014

N2 - Modeling coarticulation in speech has been largely limited to short sequences and/or limited phonetic context. We introduce a methodology for modeling both formant frequency and bandwidth in continuous speech, allowing examination of sentencelevel coarticulation. The model represents continuous trajectories as a combination of overlapping local trajectories, which are represented by a weighted-addition of acoustic event targets by sigmoidal coarticulation functions characterized by slope and position. Estimation is achieved using a combination of hill-climbing and grid-search, with global target, joint slope for identical contexts, and local position parameters. We evaluate model performance for two speakers using an intelligibility test that compares vocoded model output to a purely vocoded and a natural condition.

AB - Modeling coarticulation in speech has been largely limited to short sequences and/or limited phonetic context. We introduce a methodology for modeling both formant frequency and bandwidth in continuous speech, allowing examination of sentencelevel coarticulation. The model represents continuous trajectories as a combination of overlapping local trajectories, which are represented by a weighted-addition of acoustic event targets by sigmoidal coarticulation functions characterized by slope and position. Estimation is achieved using a combination of hill-climbing and grid-search, with global target, joint slope for identical contexts, and local position parameters. We evaluate model performance for two speakers using an intelligibility test that compares vocoded model output to a purely vocoded and a natural condition.

KW - Coarticulation

KW - Continuous speech

KW - Formants

UR - http://www.scopus.com/inward/record.url?scp=84910047777&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84910047777&partnerID=8YFLogxK

M3 - Conference contribution

SP - 193

EP - 197

BT - INTERSPEECH 2014 - 15th Annual Conference of the International Speech Communication Association

PB - International Speech and Communication Association

ER -