dc.contributor.author |
Sahraeian, R
|
|
dc.contributor.author |
Van Compernolle, D
|
|
dc.contributor.author |
De Wet, Febe
|
|
dc.date.accessioned |
2016-08-19T08:13:23Z |
|
dc.date.available |
2016-08-19T08:13:23Z |
|
dc.date.issued |
2015-11 |
|
dc.identifier.citation |
Sahraeian, R, Van Compernolle, D and De Wet, F. 2015. Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans. In: Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference (PRASA-RobMech), Port Elizabeth, South Africa, 25-26 November 2015 |
en_US |
dc.identifier.isbn |
978-1-4673-7450-7/15 |
|
dc.identifier.uri |
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7359508&tag=1
|
|
dc.identifier.uri |
http://hdl.handle.net/10204/8714
|
|
dc.description |
Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference (PRASA-RobMech), Port Elizabeth, South Africa, 25-26 November 2015. |
en_US |
dc.description.abstract |
Recently, multilingual deep neural networks (DNNs) have been successfully used to improve under-resourced speech recognizers. Common approaches use either a merged universal phoneme set based on the International Phonetic Alphabet (IPA) or a language specific phoneme set to train a multilingual DNN. In this paper, we investigate the effect of both knowledge-based and data-driven phoneme mapping on the multilingual DNN and its application to an under-resourced language. For the data-driven phoneme mapping we propose to use an approximation of Kullback Leibler Divergence (KLD) to generate a confusion matrix and find the best matching phonemes of the target language for each individual phoneme in the donor language. Moreover, we explore the use of recently proposed generalized maxout network in both multilingual and low resource monolingual scenarios. We evaluate the proposed phoneme mappings on a phoneme recognition task with both HMM/GMM and DNN systems with generalized maxout architecture where Flemish and Afrikaans are used as donor and under-resourced target languages respectively. |
en_US |
dc.language.iso |
en |
en_US |
dc.publisher |
IEEE |
en_US |
dc.relation.ispartofseries |
Workflow;16008 |
|
dc.subject |
Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference |
en_US |
dc.subject |
PRASA-RobMech |
en_US |
dc.subject |
Multilingual deep neural network |
en_US |
dc.subject |
Kullback Leibler Divergence |
en_US |
dc.subject |
Phoneme mapping |
en_US |
dc.subject |
Automatic speech recognition |
en_US |
dc.subject |
ASR |
en_US |
dc.subject |
Low resource ASR |
en_US |
dc.title |
Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans |
en_US |
dc.type |
Conference Presentation |
en_US |
dc.identifier.apacitation |
Sahraeian, R., Van Compernolle, D., & De Wet, F. (2015). Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans. IEEE. http://hdl.handle.net/10204/8714 |
en_ZA |
dc.identifier.chicagocitation |
Sahraeian, R, D Van Compernolle, and Febe De Wet. "Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans." (2015): http://hdl.handle.net/10204/8714 |
en_ZA |
dc.identifier.vancouvercitation |
Sahraeian R, Van Compernolle D, De Wet F, Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans; IEEE; 2015. http://hdl.handle.net/10204/8714 . |
en_ZA |
dc.identifier.ris |
TY - Conference Presentation
AU - Sahraeian, R
AU - Van Compernolle, D
AU - De Wet, Febe
AB - Recently, multilingual deep neural networks (DNNs) have been successfully used to improve under-resourced speech recognizers. Common approaches use either a merged universal phoneme set based on the International Phonetic Alphabet (IPA) or a language specific phoneme set to train a multilingual DNN. In this paper, we investigate the effect of both knowledge-based and data-driven phoneme mapping on the multilingual DNN and its application to an under-resourced language. For the data-driven phoneme mapping we propose to use an approximation of Kullback Leibler Divergence (KLD) to generate a confusion matrix and find the best matching phonemes of the target language for each individual phoneme in the donor language. Moreover, we explore the use of recently proposed generalized maxout network in both multilingual and low resource monolingual scenarios. We evaluate the proposed phoneme mappings on a phoneme recognition task with both HMM/GMM and DNN systems with generalized maxout architecture where Flemish and Afrikaans are used as donor and under-resourced target languages respectively.
DA - 2015-11
DB - ResearchSpace
DP - CSIR
KW - Pattern Recognition Association of South Africa and Robotics and Mechatronics International Conference
KW - PRASA-RobMech
KW - Multilingual deep neural network
KW - Kullback Leibler Divergence
KW - Phoneme mapping
KW - Automatic speech recognition
KW - ASR
KW - Low resource ASR
LK - https://researchspace.csir.co.za
PY - 2015
SM - 978-1-4673-7450-7/15
T1 - Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans
TI - Using generalized maxout networks and phoneme mapping for low resource ASR- a case study on Flemish-Afrikaans
UR - http://hdl.handle.net/10204/8714
ER -
|
en_ZA |