Consonant discrimination in elicited and spontaneous speech: A case for signal-adaptive front ends in ASR

Mustafa (Kemal) Sonmez, Madelaine Plauché, Elizabeth Shriberg, Horacio Franco

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

6 Citations (Scopus)

Abstract

The constant frame length in typical ASR front ends is too long to capture transient phenomena in speech, such as stop bursts. However, current HMM systems have consistently outperformed systems based solely on non-uniform units. This work investigates an approach to "add back" such transient information to a speech recognizer, without losing the robustness of the standard acoustic models. We demonstrate a set of phonetically-motivated acoustic features that discriminate a preliminary test set of highly ambiguous voiceless stops in CV contexts. The features are automatically computed from data that had been hand-marked for consonant burst location and voicing onset (extension to automatic marking is also proposed). Two corpora are processed using a parallel set of features: conversational speech over the telephone (Switchboard), and a corpus of carefully elicited speech. The latter provides an upper bound on discrimination, and allows for comparison of feature usage across speaking style. We explore data-driven approaches to obtaining variable-length time-localized features compatible with an HMM statistical framework. We also suggest techniques for extension to automatic annotation of burst location, for computation of features at such points, and for augmentation of an HMM system with the added information.

Original languageEnglish (US)
Title of host publication6th International Conference on Spoken Language Processing, ICSLP 2000
PublisherInternational Speech Communication Association
ISBN (Electronic)7801501144, 9787801501141
StatePublished - 2000
Externally publishedYes
Event6th International Conference on Spoken Language Processing, ICSLP 2000 - Beijing, China
Duration: Oct 16 2000Oct 20 2000

Other

Other6th International Conference on Spoken Language Processing, ICSLP 2000
CountryChina
CityBeijing
Period10/16/0010/20/00

Fingerprint

discrimination
acoustics
telephone
speaking
Consonant
Spontaneous Speech
Discrimination
Hidden Markov Model
Length
Acoustics

ASJC Scopus subject areas

  • Linguistics and Language
  • Language and Linguistics

Cite this

Sonmez, M. K., Plauché, M., Shriberg, E., & Franco, H. (2000). Consonant discrimination in elicited and spontaneous speech: A case for signal-adaptive front ends in ASR. In 6th International Conference on Spoken Language Processing, ICSLP 2000 International Speech Communication Association.

Consonant discrimination in elicited and spontaneous speech : A case for signal-adaptive front ends in ASR. / Sonmez, Mustafa (Kemal); Plauché, Madelaine; Shriberg, Elizabeth; Franco, Horacio.

6th International Conference on Spoken Language Processing, ICSLP 2000. International Speech Communication Association, 2000.

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

Sonmez, MK, Plauché, M, Shriberg, E & Franco, H 2000, Consonant discrimination in elicited and spontaneous speech: A case for signal-adaptive front ends in ASR. in 6th International Conference on Spoken Language Processing, ICSLP 2000. International Speech Communication Association, 6th International Conference on Spoken Language Processing, ICSLP 2000, Beijing, China, 10/16/00.
Sonmez MK, Plauché M, Shriberg E, Franco H. Consonant discrimination in elicited and spontaneous speech: A case for signal-adaptive front ends in ASR. In 6th International Conference on Spoken Language Processing, ICSLP 2000. International Speech Communication Association. 2000
Sonmez, Mustafa (Kemal) ; Plauché, Madelaine ; Shriberg, Elizabeth ; Franco, Horacio. / Consonant discrimination in elicited and spontaneous speech : A case for signal-adaptive front ends in ASR. 6th International Conference on Spoken Language Processing, ICSLP 2000. International Speech Communication Association, 2000.
@inproceedings{235dba2c24b7455280feabd4e41c9e89,
title = "Consonant discrimination in elicited and spontaneous speech: A case for signal-adaptive front ends in ASR",
abstract = "The constant frame length in typical ASR front ends is too long to capture transient phenomena in speech, such as stop bursts. However, current HMM systems have consistently outperformed systems based solely on non-uniform units. This work investigates an approach to {"}add back{"} such transient information to a speech recognizer, without losing the robustness of the standard acoustic models. We demonstrate a set of phonetically-motivated acoustic features that discriminate a preliminary test set of highly ambiguous voiceless stops in CV contexts. The features are automatically computed from data that had been hand-marked for consonant burst location and voicing onset (extension to automatic marking is also proposed). Two corpora are processed using a parallel set of features: conversational speech over the telephone (Switchboard), and a corpus of carefully elicited speech. The latter provides an upper bound on discrimination, and allows for comparison of feature usage across speaking style. We explore data-driven approaches to obtaining variable-length time-localized features compatible with an HMM statistical framework. We also suggest techniques for extension to automatic annotation of burst location, for computation of features at such points, and for augmentation of an HMM system with the added information.",
author = "Sonmez, {Mustafa (Kemal)} and Madelaine Plauch{\'e} and Elizabeth Shriberg and Horacio Franco",
year = "2000",
language = "English (US)",
booktitle = "6th International Conference on Spoken Language Processing, ICSLP 2000",
publisher = "International Speech Communication Association",

}

TY - GEN

T1 - Consonant discrimination in elicited and spontaneous speech

T2 - A case for signal-adaptive front ends in ASR

AU - Sonmez, Mustafa (Kemal)

AU - Plauché, Madelaine

AU - Shriberg, Elizabeth

AU - Franco, Horacio

PY - 2000

Y1 - 2000

N2 - The constant frame length in typical ASR front ends is too long to capture transient phenomena in speech, such as stop bursts. However, current HMM systems have consistently outperformed systems based solely on non-uniform units. This work investigates an approach to "add back" such transient information to a speech recognizer, without losing the robustness of the standard acoustic models. We demonstrate a set of phonetically-motivated acoustic features that discriminate a preliminary test set of highly ambiguous voiceless stops in CV contexts. The features are automatically computed from data that had been hand-marked for consonant burst location and voicing onset (extension to automatic marking is also proposed). Two corpora are processed using a parallel set of features: conversational speech over the telephone (Switchboard), and a corpus of carefully elicited speech. The latter provides an upper bound on discrimination, and allows for comparison of feature usage across speaking style. We explore data-driven approaches to obtaining variable-length time-localized features compatible with an HMM statistical framework. We also suggest techniques for extension to automatic annotation of burst location, for computation of features at such points, and for augmentation of an HMM system with the added information.

AB - The constant frame length in typical ASR front ends is too long to capture transient phenomena in speech, such as stop bursts. However, current HMM systems have consistently outperformed systems based solely on non-uniform units. This work investigates an approach to "add back" such transient information to a speech recognizer, without losing the robustness of the standard acoustic models. We demonstrate a set of phonetically-motivated acoustic features that discriminate a preliminary test set of highly ambiguous voiceless stops in CV contexts. The features are automatically computed from data that had been hand-marked for consonant burst location and voicing onset (extension to automatic marking is also proposed). Two corpora are processed using a parallel set of features: conversational speech over the telephone (Switchboard), and a corpus of carefully elicited speech. The latter provides an upper bound on discrimination, and allows for comparison of feature usage across speaking style. We explore data-driven approaches to obtaining variable-length time-localized features compatible with an HMM statistical framework. We also suggest techniques for extension to automatic annotation of burst location, for computation of features at such points, and for augmentation of an HMM system with the added information.

UR - http://www.scopus.com/inward/record.url?scp=85009115694&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85009115694&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:85009115694

BT - 6th International Conference on Spoken Language Processing, ICSLP 2000

PB - International Speech Communication Association

ER -