Adaptation of Acoustic Models in Joint Speaker and Noise Space Using Bilinear Models

Yongwon JEONG; Hyung Soon KIM

doi:10.1587/transinf.E97.D.2195

Adaptation of Acoustic Models in Joint Speaker and Noise Space Using Bilinear Models

Yongwon JEONG, Hyung Soon KIM

Full Text Views

0

Share
Cite this

Summary :

We present the adaptation of the acoustic models of hidden Markov models (HMMs) to the target speaker and noise environment using bilinear models. Acoustic models trained from various speakers and noise conditions are decomposed to build the bases that capture the interaction between the two factors. The model for the target speaker and noise is represented as a product of bases and two weight vectors. In experiments using the AURORA4 corpus, the bilinear model outperforms the linear model.

Publication: IEICE TRANSACTIONS on Information Vol.E97-D No.8 pp.2195-2199

Publication Date: 2014/08/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.E97.D.2195

Type of Manuscript: LETTER

Category: Speech and Hearing

Authors

Yongwon JEONG
Pusan National University
Hyung Soon KIM
Pusan National University

Keyword

bilinear model, eigenvoice speaker adaptation, environment adaptation, speaker adaptation, speech recognition

Cite this

Copy

Yongwon JEONG, Hyung Soon KIM, "Adaptation of Acoustic Models in Joint Speaker and Noise Space Using Bilinear Models" in IEICE TRANSACTIONS on Information, vol. E97-D, no. 8, pp. 2195-2199, August 2014, doi: 10.1587/transinf.E97.D.2195.
Abstract: We present the adaptation of the acoustic models of hidden Markov models (HMMs) to the target speaker and noise environment using bilinear models. Acoustic models trained from various speakers and noise conditions are decomposed to build the bases that capture the interaction between the two factors. The model for the target speaker and noise is represented as a product of bases and two weight vectors. In experiments using the AURORA4 corpus, the bilinear model outperforms the linear model.
URL: https://globals.ieice.org/en_transactions/information/10.1587/transinf.E97.D.2195/_p

Copy

@ARTICLE{e97-d_8_2195,
author={Yongwon JEONG, Hyung Soon KIM, },
journal={IEICE TRANSACTIONS on Information},
title={Adaptation of Acoustic Models in Joint Speaker and Noise Space Using Bilinear Models},
year={2014},
volume={E97-D},
number={8},
pages={2195-2199},
abstract={We present the adaptation of the acoustic models of hidden Markov models (HMMs) to the target speaker and noise environment using bilinear models. Acoustic models trained from various speakers and noise conditions are decomposed to build the bases that capture the interaction between the two factors. The model for the target speaker and noise is represented as a product of bases and two weight vectors. In experiments using the AURORA4 corpus, the bilinear model outperforms the linear model.},
keywords={},
doi={10.1587/transinf.E97.D.2195},
ISSN={1745-1361},
month={August},}

Copy

TY - JOUR
TI - Adaptation of Acoustic Models in Joint Speaker and Noise Space Using Bilinear Models
T2 - IEICE TRANSACTIONS on Information
SP - 2195
EP - 2199
AU - Yongwon JEONG
AU - Hyung Soon KIM
PY - 2014
DO - 10.1587/transinf.E97.D.2195
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E97-D
IS - 8
JA - IEICE TRANSACTIONS on Information
Y1 - August 2014
AB - We present the adaptation of the acoustic models of hidden Markov models (HMMs) to the target speaker and noise environment using bilinear models. Acoustic models trained from various speakers and noise conditions are decomposed to build the bases that capture the interaction between the two factors. The model for the target speaker and noise is represented as a product of bases and two weight vectors. In experiments using the AURORA4 corpus, the bilinear model outperforms the linear model.
ER -