Click to open the HelpDesk interface
AECE - Front page banner

Menu:


FACTS & FIGURES

JCR Impact Factor: 0.459
JCR 5-Year IF: 0.442
Issues per year: 4
Current issue: Nov 2016
Next issue: Feb 2017
Avg review time: 78 days


PUBLISHER

Stefan cel Mare
University of Suceava
Faculty of Electrical Engineering and
Computer Science
13, Universitatii Street
Suceava - 720229
ROMANIA

Print ISSN: 1582-7445
Online ISSN: 1844-7600
WorldCat: 643243560
doi: 10.4316/AECE


TRAFFIC STATS

1,530,072 unique visits
477,939 downloads
Since November 1, 2009



No robots online now


SJR SCImago RANK

SCImago Journal & Country Rank


SEARCH ENGINES

aece.ro - Google Pagerank




TEXT LINKS

Anycast DNS Hosting
MOST RECENT ISSUES

 Volume 16 (2016)
 
     »   Issue 4 / 2016
 
     »   Issue 3 / 2016
 
     »   Issue 2 / 2016
 
     »   Issue 1 / 2016
 
 
 Volume 15 (2015)
 
     »   Issue 4 / 2015
 
     »   Issue 3 / 2015
 
     »   Issue 2 / 2015
 
     »   Issue 1 / 2015
 
 
 Volume 14 (2014)
 
     »   Issue 4 / 2014
 
     »   Issue 3 / 2014
 
     »   Issue 2 / 2014
 
     »   Issue 1 / 2014
 
 
 Volume 13 (2013)
 
     »   Issue 4 / 2013
 
     »   Issue 3 / 2013
 
     »   Issue 2 / 2013
 
     »   Issue 1 / 2013
 
 
  View all issues  


FEATURED ARTICLE

Broken Bar Fault Detection in IM Operating Under No-Load Condition, RELJIC, D., JERKAN, D., MARCETIC, D., OROS, D.
Issue 4/2016

AbstractPlus






LATEST NEWS

2017-Feb-16
With new technologies, such as mobile communications, internet of things, and wide applications of social media, organizations generate a huge volume of data, much faster than several years ago. Big data, characterized by high volume, diversity and velocity, increasingly drives decision making and is changing the landscape of business intelligence, from governments to private organizations, from communities to individuals. Big data analytics that discover insights from evidences has a high demand for computing efficiency, knowledge discovery, problem solving, and event prediction. We dedicate a special section of Issue 4/2017 to Big Data. Prospective authors are asked to make the submissions for this section no later than the 31st of May 2017, placing "Big Data - " before the paper title in OpenConf.

2017-Jan-30
We have the confirmation Advances in Electrical and Computer Engineering will be included in the Gale database.

2016-Dec-17
IoT is a new emerging technology domain which will be used to connect all objects through the Internet for remote sensing and control. IoT uses a combination of WSN (Wireless Sensor Network), M2M (Machine to Machine), robotics, wireless networking, Internet technologies, and Smart Devices. We dedicate a special section of Issue 2/2017 to IoT. Prospective authors are asked to make the submissions for this section no later than the 31st of March 2017, placing "IoT - " before the paper title in OpenConf.

2016-Jun-14
Thomson Reuters published the Journal Citations Report for 2015. The JCR Impact Factor of Advances in Electrical and Computer Engineering is 0.459, and the JCR 5-Year Impact Factor is 0.442.

2015-Dec-04
Starting with Issue 2/2016, the article processing charge is 300 EUR for each article accepted for publication. The charge of 25 EUR per page for papers over 8 pages will not be changed. Details are available in the For authors section.

Read More »


    
 

  1/2015 - 9

Enhancing ASR Systems for Under-Resourced Languages through a Novel Unsupervised Acoustic Model Training Technique

CUCU, H. See more information about CUCU, H. on SCOPUS See more information about CUCU, H. on IEEExplore See more information about CUCU, H. on Web of Science, BUZO, A. See more information about  BUZO, A. on SCOPUS See more information about  BUZO, A. on SCOPUS See more information about BUZO, A. on Web of Science, BESACIER, L. See more information about  BESACIER, L. on SCOPUS See more information about  BESACIER, L. on SCOPUS See more information about BESACIER, L. on Web of Science, BURILEANU, C. See more information about BURILEANU, C. on SCOPUS See more information about BURILEANU, C. on SCOPUS See more information about BURILEANU, C. on Web of Science
 
Click to see author's profile on See more information about the author on SCOPUS SCOPUS, See more information about the author on IEEE Xplore IEEE Xplore, See more information about the author on Web of Science Web of Science

Download PDF pdficon (612 KB) | Citation | Downloads: 205 | Views: 1,281

Author keywords
speech recognition, under-resourced languages, unsupervised acoustic modeling, unsupervised training

References keywords
speech(15), training(13), unsupervised(12), resourced(5), recognition(5), processing(5), languages(5), language(5), acoustic(5), system(4)
Blue keywords are present in both the references section and the paper title.

About this article
Date of Publication: 2015-02-28
Volume 15, Issue 1, Year 2015, On page(s): 63 - 68
ISSN: 1582-7445, e-ISSN: 1844-7600
Digital Object Identifier: 10.4316/AECE.2015.01009
Web of Science Accession Number: 000352158600009
SCOPUS ID: 84924787729

Abstract
Quick view
Full text preview
Statistical speech and language processing techniques, requiring large amounts of training data, are currently state-of-the-art in automatic speech recognition. For high-resourced, international languages this data is widely available, while for under-resourced languages the lack of data poses serious problems. Unsupervised acoustic modeling can offer a cost and time effective way of creating a solid acoustic model for any under-resourced language. This study describes a novel unsupervised acoustic model training method and evaluates it on speech data in an under-resourced language: Romanian. The key novel factor of the method is the usage of two complementary seed ASR systems to produce high quality transcriptions, with a Character Error Rate (ChER) < 5%, for initially untranscribed speech data. The methodology leads to a relative Word Error Rate (WER) improvement of more than 10% when 100 hours of untranscribed speech are used.


References | Cited By  «-- Click to see who has cited this paper

[1] L. Besacier, E. Barnard, A. Karpov, T. Schultz, "Automatic speech recognition for under-resourced languages: A survey.", in Speech Communication, Vol. 56 - Special Issue on Processing Under-Resourced Languages, pp. 85-100,
[CrossRef] [Web of Science Times Cited 30] [SCOPUS Times Cited 50]


[2] H. Cucu, "Towards a speaker-independent, large-vocabulary continuous speech recognition system for Romanian", PhD Thesis, University "Politehnica" of Bucharest, 2011.

[3] A. Buzo, H. Cucu, C. Burileanu, "Text Spotting In Large Speech Databases For Under-Resourced Languages", in Proc. Int. Conf. Speech Technology and Human-Computer Dialogue (SpeD), Cluj-Napoca, Romania, 2013, pp. 77-82,
[CrossRef] [SCOPUS Record]


[4] H. Cucu, A. Buzo, C. Burileanu, "Unsupervised Acoustic Model Training using Multiple Seed ASR Systems", in Proc. Int. Workshop on Spoken Language Technologies for Under-resourced Languages (SLTU), St. Petersburg, Russia, 2014, pp. 124-130.

[5] G. Zavaliagkos, T. Colthurst, "Utilizing Untranscribed Training Data to Improve Performance", in DARPA Broadcast News Transcription and Understanding Workshop, Lansdowne, USA, 1998, pp. 301-305

[6] T. Kemp and A. Waibel, "Unsupervised Training of a Speech Recognizer: Recent Experiments", in Proc. Eurospeech, Budapest, Hungary, 1999, pp. 2725-2728.

[7] F. Wessel and H. Ney, "Unsupervised training of acoustic models for large vocabulary continuous speech recognition", in Proc. Automatic Speech Recognition and Understanding Workshop (ASRU), Trento, Italy, 2001, pp. 307-310,
[CrossRef] [Web of Science Times Cited 32] [SCOPUS Times Cited 49]


[8] L. Lamel, J.-L. Gauvain, G. Adda, "Lightly Supervised and Unsupervised Acoustic Model Training", in Computer Speech & Language, vol. 16, pp. 115-129, 2002. Available:
[CrossRef] [Web of Science Times Cited 96] [SCOPUS Times Cited 135]


[9] T. Fraga-Silva, J.-L. Gauvain, L. Lamel, "Lattice-based Unsupervised Acoustic Model Training", in Proc. Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), Prague, Czech Republic, 2011, pp. 4656-4659,
[CrossRef] [SCOPUS Times Cited 12]


[10] L. Wang, M. J. F. Gales and P. C. Woodland, "Unsupervised training for mandarin broadcast news and conversational transcription", in Proc. Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), Honolulu, Hawaii, 2007, vol. IV, pp. 353-356,
[CrossRef] [SCOPUS Times Cited 16]


[11] J. Ma, S. Matsoukas., "Unsupervised training on a large amount of Arabic news broadcast data", in Proc. Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), Hawaii, 2007, vol. II, pp. 349-352,
[CrossRef] [SCOPUS Times Cited 1]


[12] K. Yu, M. J. F. Gales, L. Wang and P. C. Woodland, "Unsupervised training and directed manual transcription for LVCSR", in Speech Communication, Vol. 52, pp. 652-663, 2010. Available:
[CrossRef] [Web of Science Times Cited 14] [SCOPUS Times Cited 23]


[13] J. Loof, C. Gollan, and H. Ney, "Cross-language Bootstrapping for Unsupervised Acoustic Model Training: Rapid Development of a Polish Speech Recognition System", in Proc. INTERSPEECH, Brighton, U.K., 2009, pp. 88-91.

[14] N. T. Vu, F. Kraus and T. Schultz, "Cross-language bootstrapping based on completely unsupervised training using multilingual A-stabil", in Proc. Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), Prague, Czech Republic, 2011, pp. 5000-5003,
[CrossRef] [SCOPUS Times Cited 15]


[15] N. T. Vu, F. Kraus and T. Schultz, "Rapid building of an ASR system for Under-Resourced Languages based on Multilingual Unsupervised Training", In Proc. INTERSPEECH, Florence, Italy, 2011, pp. 3145-3148.

[16] N. T. Vu, F. Kraus and T. Schultz, "Multilingual A-stabil: A new confidence score for multilingual unsupervised training", in Spoken Language Technology Workshop (SLT), Berkeley, California, USA, 2010, pp. 183-188,
[CrossRef] [SCOPUS Times Cited 5]


[17] H. Cucu, A. Buzo, L. Petrica, D. Burileanu and C. Burileanu, "Recent Improvements of the SpeeD Romanian LVCSR System", in Proc. Int. Conf. on Communications (COMM), Bucharest, Romania, 2014, pp. 111-114,
[CrossRef] [SCOPUS Times Cited 3]


[18] CMU Sphinx Toolkit: [Online] Available: Temporary on-line reference link removed - see the PDF document

[19] SRI-LM Toolkit: [Online] Available: Temporary on-line reference link removed - see the PDF document

[20] M. Rouvier, G. Dupuy, P. Gay, E. Khoury, T. Merlin, S. Meignier, "An Open-source State-of-the-art Toolbox for Broadcast News Diarization," in Proc. INTERSPEECH, Lyon, France, 2013.



References Weight

Web of Science® Citations for all references: 172 TCR
SCOPUS® Citations for all references: 309 TCR

Web of Science® Average Citations per reference: 8 ACR
SCOPUS® Average Citations per reference: 15 ACR

TCR = Total Citations for References / ACR = Average Citations per Reference

We introduced in 2010 - for the first time in scientific publishing, the term "References Weight", as a quantitative indication of the quality ... Read more

Citations for references background updated on 2017-02-18 11:31 in 68 seconds.




Note1: Web of Science® is a registered trademark of Thomson Reuters.
Note2: SCOPUS® is a registered trademark of Elsevier B.V.
Disclaimer: All queries to the respective databases were made by using the DOI record of every reference (where available). Due to technical problems beyond our control, the information is not always accurate. Please use the CrossRef link to visit the respective publisher site.

Copyright ©2001-2017
Faculty of Electrical Engineering and Computer Science
Stefan cel Mare University of Suceava, Romania


All rights reserved: Advances in Electrical and Computer Engineering is a registered trademark of the Stefan cel Mare University of Suceava. No part of this publication may be reproduced, stored in a retrieval system, photocopied, recorded or archived, without the written permission from the Editor. When authors submit their papers for publication, they agree that the copyright for their article be transferred to the Faculty of Electrical Engineering and Computer Science, Stefan cel Mare University of Suceava, Romania, if and only if the articles are accepted for publication. The copyright covers the exclusive rights to reproduce and distribute the article, including reprints and translations.

Permission for other use: The copyright owner's consent does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific written permission must be obtained from the Editor for such copying. Direct linking to files hosted on this website is strictly prohibited.

Disclaimer: Whilst every effort is made by the publishers and editorial board to see that no inaccurate or misleading data, opinions or statements appear in this journal, they wish to make it clear that all information and opinions formulated in the articles, as well as linguistic accuracy, are the sole responsibility of the author.




Website loading speed and performance optimization powered by: