Cover Image

Merging of native and non-native speech for low-resource accented ASR

This paper presents our recent study on low-resource automatic speech recognition (ASR) system with accented speech. We propose multi-accent Subspace Gaussian Mixture Models (SGMM) and accent-specific Deep Neural Networks (DNN) for improving non-native ASR performance. In the SGMM framework, we pres...

Full description

Saved in:
Bibliographic Details
Main Authors: Samson Juan, Sarah, Besacier, Laurent, Lecouteux, Benjamin, Tien-Ping, Tan
Format: E-Article
Language:English
Published: Springer Verlag 2015
Subjects:
T Technology (General)
Online Access:http://ir.unimas.my/id/eprint/12098/1/No%2035%20%28abstrak%29.pdf
http://ir.unimas.my/id/eprint/12098/
http://www.scopus.com/inward/record.url?eid=2-s2.0-84952362047&partnerID=40&md5=6bc512988afc29cd7ca4af16a836f0b3
Tags: Add Tag
No Tags, Be the first to tag this record!
id my.unimas.ir.12098
record_format eprints
spelling my.unimas.ir.120982016-10-21T07:34:47Z http://ir.unimas.my/id/eprint/12098/ Merging of native and non-native speech for low-resource accented ASR Samson Juan, Sarah Besacier, Laurent Lecouteux, Benjamin Tien-Ping, Tan T Technology (General) This paper presents our recent study on low-resource automatic speech recognition (ASR) system with accented speech. We propose multi-accent Subspace Gaussian Mixture Models (SGMM) and accent-specific Deep Neural Networks (DNN) for improving non-native ASR performance. In the SGMM framework, we present an original language weighting strategy to merge the globally shared parameters of two models based on native and non-native speech espectively. In the DNN framework, a native deep neural net is fine-tuned to non-native speech. Over the non-native baseline, we achieved relative improvement of 15% for multi-accent SGMM and 34% for accent-specific DNN with speaker adaptation. Springer Verlag 2015 E-Article PeerReviewed text en http://ir.unimas.my/id/eprint/12098/1/No%2035%20%28abstrak%29.pdf Samson Juan, Sarah and Besacier, Laurent and Lecouteux, Benjamin and Tien-Ping, Tan (2015) Merging of native and non-native speech for low-resource accented ASR. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 9449. pp. 255-266. ISSN 3029743 http://www.scopus.com/inward/record.url?eid=2-s2.0-84952362047&partnerID=40&md5=6bc512988afc29cd7ca4af16a836f0b3 DOI: 10.1007/978-3-319-25789-1 24
institution Universiti Malaysia Sarawak
building Centre for Academic Information Services (CAIS)
collection Institutional Repository
continent Asia
country Malaysia
content_provider Universiti Malaysia Sarawak
content_source UNIMAS Institutional Repository
url_provider http://ir.unimas.my/
language English
topic T Technology (General)
spellingShingle T Technology (General)
Samson Juan, Sarah
Besacier, Laurent
Lecouteux, Benjamin
Tien-Ping, Tan
Merging of native and non-native speech for low-resource accented ASR
description This paper presents our recent study on low-resource automatic speech recognition (ASR) system with accented speech. We propose multi-accent Subspace Gaussian Mixture Models (SGMM) and accent-specific Deep Neural Networks (DNN) for improving non-native ASR performance. In the SGMM framework, we present an original language weighting strategy to merge the globally shared parameters of two models based on native and non-native speech espectively. In the DNN framework, a native deep neural net is fine-tuned to non-native speech. Over the non-native baseline, we achieved relative improvement of 15% for multi-accent SGMM and 34% for accent-specific DNN with speaker adaptation.
format E-Article
author Samson Juan, Sarah
Besacier, Laurent
Lecouteux, Benjamin
Tien-Ping, Tan
author_facet Samson Juan, Sarah
Besacier, Laurent
Lecouteux, Benjamin
Tien-Ping, Tan
author_sort Samson Juan, Sarah
title Merging of native and non-native speech for low-resource accented ASR
title_short Merging of native and non-native speech for low-resource accented ASR
title_full Merging of native and non-native speech for low-resource accented ASR
title_fullStr Merging of native and non-native speech for low-resource accented ASR
title_full_unstemmed Merging of native and non-native speech for low-resource accented ASR
title_sort merging of native and non-native speech for low-resource accented asr
publisher Springer Verlag
publishDate 2015
url http://ir.unimas.my/id/eprint/12098/1/No%2035%20%28abstrak%29.pdf
http://ir.unimas.my/id/eprint/12098/
http://www.scopus.com/inward/record.url?eid=2-s2.0-84952362047&partnerID=40&md5=6bc512988afc29cd7ca4af16a836f0b3
_version_ 1644511342444412928
score 13.211869

Similar Items