Improving Children's Speech Recognition through Out-of-Domain Data Augmentation

Joachim Fainberg; Peter Bell; Mike Lincoln; Steve Renals

doi:10.21437/Interspeech.2016-1348

Improving Children's Speech Recognition through Out-of-Domain Data Augmentation

Joachim Fainberg, Peter Bell, Mike Lincoln, Steve Renals

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Children's speech poses challenges to speech recognition due to strong age-dependent anatomical variations and a lack of large, publicly-available corpora. In this paper we explore data augmentation for children's speech recognition using stochastic feature mapping (SFM) to transform out-of-domain adult data for both GMM-based and DNN-based acoustic models. We performed experiments on the English PF-STAR corpus, augmenting using WSJCAM0 and ABI. Our experimental results indicate that a DNN acoustic model for childrens speech can make use of adult data, and that out-of-domain SFM is more accurate than in-domain SFM.

Original language	English
Title of host publication	Interspeech 2016
Pages	1598-1602
Number of pages	5
DOIs	https://doi.org/10.21437/Interspeech.2016-1348
Publication status	Published - 12 Sept 2016
Event	Interspeech 2016 - San Francisco, United States Duration: 8 Sept 2016 → 12 Sept 2016 http://www.interspeech2016.org/

Publication series

Name	Interspeech
Publisher	International Speech Communication Association
ISSN (Print)	1990-9772

Conference

Conference	Interspeech 2016
Country/Territory	United States
City	San Francisco
Period	8/09/16 → 12/09/16
Internet address	http://www.interspeech2016.org/

Access to Document

10.21437/Interspeech.2016-1348

jfainberg-is2016Accepted author manuscript, 214 KB

http://www.isca-speech.org/archive/Interspeech_2016/abstracts/1348.html

Multi-domain speech recognition
Renals, S. (Principal Investigator)
Non-EU industry, commerce and public corporations
1/09/15 → 28/02/19
Project: Research

Cite this

@inproceedings{05600b7caff14c9da5a8934a511f4fe5,

title = "Improving Children's Speech Recognition through Out-of-Domain Data Augmentation",

abstract = "Children's speech poses challenges to speech recognition due to strong age-dependent anatomical variations and a lack of large, publicly-available corpora. In this paper we explore data augmentation for children's speech recognition using stochastic feature mapping (SFM) to transform out-of-domain adult data for both GMM-based and DNN-based acoustic models. We performed experiments on the English PF-STAR corpus, augmenting using WSJCAM0 and ABI. Our experimental results indicate that a DNN acoustic model for childrens speech can make use of adult data, and that out-of-domain SFM is more accurate than in-domain SFM.",

author = "Joachim Fainberg and Peter Bell and Mike Lincoln and Steve Renals",

year = "2016",

month = sep,

day = "12",

doi = "10.21437/Interspeech.2016-1348",

language = "English",

series = "Interspeech",

publisher = "International Speech Communication Association",

pages = "1598--1602",

booktitle = "Interspeech 2016",

note = "Interspeech 2016 ; Conference date: 08-09-2016 Through 12-09-2016",

url = "http://www.interspeech2016.org/",

}

TY - GEN

T1 - Improving Children's Speech Recognition through Out-of-Domain Data Augmentation

AU - Fainberg, Joachim

AU - Bell, Peter

AU - Lincoln, Mike

AU - Renals, Steve

PY - 2016/9/12

Y1 - 2016/9/12

N2 - Children's speech poses challenges to speech recognition due to strong age-dependent anatomical variations and a lack of large, publicly-available corpora. In this paper we explore data augmentation for children's speech recognition using stochastic feature mapping (SFM) to transform out-of-domain adult data for both GMM-based and DNN-based acoustic models. We performed experiments on the English PF-STAR corpus, augmenting using WSJCAM0 and ABI. Our experimental results indicate that a DNN acoustic model for childrens speech can make use of adult data, and that out-of-domain SFM is more accurate than in-domain SFM.

AB - Children's speech poses challenges to speech recognition due to strong age-dependent anatomical variations and a lack of large, publicly-available corpora. In this paper we explore data augmentation for children's speech recognition using stochastic feature mapping (SFM) to transform out-of-domain adult data for both GMM-based and DNN-based acoustic models. We performed experiments on the English PF-STAR corpus, augmenting using WSJCAM0 and ABI. Our experimental results indicate that a DNN acoustic model for childrens speech can make use of adult data, and that out-of-domain SFM is more accurate than in-domain SFM.

U2 - 10.21437/Interspeech.2016-1348

DO - 10.21437/Interspeech.2016-1348

M3 - Conference contribution

T3 - Interspeech

SP - 1598

EP - 1602

BT - Interspeech 2016

T2 - Interspeech 2016

Y2 - 8 September 2016 through 12 September 2016

ER -

Improving Children's Speech Recognition through Out-of-Domain Data Augmentation

Abstract

Publication series

Conference

Access to Document

Fingerprint

Projects

Multi-domain speech recognition

Cite this