Information: Join today and make your research activities more affordable! Technical workshop participation fees and annual registration fees are available at member rates.
Notice: [Important] Announcement of Changes to Registration Fee Payment and Manuscript Upload Procedures for IEICE Technical Meetings
IEICE Technical Committee Submission System
Conference Paper's Information		 Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2015-08-21 16:15
Training Data Selection for Acoustic Modeling Based on Submodular Optimization of Joint KL Divergence
Taichi Asami, Ryo Masumura, Hirokazu Masataki, Manabu Okamoto, Sumitaka Sakauchi (NTT) SP2015-58
Abstract (in Japanese) (See Japanese page) 
(in English) This paper provides a novel training data selection method to
construct acoustic models for automatic speech recognition (ASR).
To deal with application-specific acoustic environments, various
training data sets have been developed for acoustic modeling.
A mixture of such already-created training sets (an large-scale set)
becomes a large utterance set containing various acoustic characteristics.
The proposed method selects the most appropriate subset of the
large-scale set and uses it for supervised training of an acoustic
model for a new ASR application.
The subset that has the most similar acoustic characteristics to the
target set (i.e. utterances recorded by the target application) is
selected based on the joint KL divergence of speech and non-speech
characteristics.
Furthermore, in order to select one of the many subsets in practical
computation time, we also propose a selection algorithm based on
submodular optimization that minimizes the joint KL divergence by
greedy selection with guaranteed optimality.
Experiments on real meeting utterances show that the proposed method
yields better acoustic models.
Keyword (in Japanese) (See Japanese page) 
(in English) speech recognition / acoustic model / training data selection / KL divergence / submodular optimization / / /  
Reference Info. IEICE Tech. Rep., vol. 115, no. 184, SP2015-58, pp. 45-50, Aug. 2015.
Paper # SP2015-58 
Date of Issue 2015-08-14 (SP) 
ISSN Print edition: ISSN 0913-5685    Online edition: ISSN 2432-6380
Copyright
and
reproduction		 All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF SP2015-58

Conference Information
Committee SP  
Conference Date 2015-08-21 - 2015-08-21 
Place (in Japanese) (See Japanese page) 
Place (in English) Iwate Prefectural Univ. 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Spoken document processing, etc. 
Paper Information
Registration To SP 
Conference Code 2015-08-SP 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Training Data Selection for Acoustic Modeling Based on Submodular Optimization of Joint KL Divergence 
Sub Title (in English)  
Keyword(1) speech recognition  
Keyword(2) acoustic model  
Keyword(3) training data selection  
Keyword(4) KL divergence  
Keyword(5) submodular optimization  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Taichi Asami  
1st Author's Affiliation Nippon Telegraph and Telephone Corporation (NTT)
2nd Author's Name Ryo Masumura  
2nd Author's Affiliation Nippon Telegraph and Telephone Corporation (NTT)
3rd Author's Name Hirokazu Masataki  
3rd Author's Affiliation Nippon Telegraph and Telephone Corporation (NTT)
4th Author's Name Manabu Okamoto  
4th Author's Affiliation Nippon Telegraph and Telephone Corporation (NTT)
5th Author's Name Sumitaka Sakauchi  
5th Author's Affiliation Nippon Telegraph and Telephone Corporation (NTT)
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
21st Author's Name  
21st Author's Affiliation ()
22nd Author's Name  
22nd Author's Affiliation ()
23rd Author's Name  
23rd Author's Affiliation ()
24th Author's Name  
24th Author's Affiliation ()
25th Author's Name  
25th Author's Affiliation ()
26th Author's Name / /
26th Author's Affiliation ()
()
27th Author's Name / /
27th Author's Affiliation ()
()
28th Author's Name / /
28th Author's Affiliation ()
()
29th Author's Name / /
29th Author's Affiliation ()
()
30th Author's Name / /
30th Author's Affiliation ()
()
31st Author's Name / /
31st Author's Affiliation ()
()
32nd Author's Name / /
32nd Author's Affiliation ()
()
33rd Author's Name / /
33rd Author's Affiliation ()
()
34th Author's Name / /
34th Author's Affiliation ()
()
35th Author's Name / /
35th Author's Affiliation ()
()
36th Author's Name / /
36th Author's Affiliation ()
()
Speaker Author-1 
Date Time 2015-08-21 16:15:00 
Presentation Time 25 minutes 
Registration for SP 
Paper # SP2015-58 
Volume (vol) vol.115 
Number (no) no.184 
Page pp.45-50 
#Pages 6 
Date of Issue 2015-08-14 (SP) 

[Return to Top Page]

[Return to IEICE Web Page]
The Institute of Electronics, Information and Communication Engineers (IEICE), Japan