The GMM-UBM framework has been proved to be one of the most effective approaches to the automatic speaker verification (ASV) task in recent years. In this letter, we first propose an approximate decision function of traditional GMM-UBM, from which it is shown that the contribution to classification of each Gaussian component is equally important. However, research in speaker perception shows that a different speech sound unit defined by Gaussian component makes a different contribution to speaker verification. This motivates us to emphasize some sound units which have discriminability between speakers while de-emphasize the speech sound units which contain little information for speaker verification. Experiments on 2006 NIST SRE core task show that the proposed approach outperforms traditional GMM-UBM approach in classification accuracy.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Xiang XIAO, Xiang ZHANG, Haipeng WANG, Hongbin SUO, Qingwei ZHAO, Yonghong YAN, "Approximate Decision Function and Optimization for GMM-UBM Based Speaker Verification" in IEICE TRANSACTIONS on Information,
vol. E92-D, no. 9, pp. 1798-1802, September 2009, doi: 10.1587/transinf.E92.D.1798.
Abstract: The GMM-UBM framework has been proved to be one of the most effective approaches to the automatic speaker verification (ASV) task in recent years. In this letter, we first propose an approximate decision function of traditional GMM-UBM, from which it is shown that the contribution to classification of each Gaussian component is equally important. However, research in speaker perception shows that a different speech sound unit defined by Gaussian component makes a different contribution to speaker verification. This motivates us to emphasize some sound units which have discriminability between speakers while de-emphasize the speech sound units which contain little information for speaker verification. Experiments on 2006 NIST SRE core task show that the proposed approach outperforms traditional GMM-UBM approach in classification accuracy.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E92.D.1798/_p
Copy
@ARTICLE{e92-d_9_1798,
author={Xiang XIAO, Xiang ZHANG, Haipeng WANG, Hongbin SUO, Qingwei ZHAO, Yonghong YAN, },
journal={IEICE TRANSACTIONS on Information},
title={Approximate Decision Function and Optimization for GMM-UBM Based Speaker Verification},
year={2009},
volume={E92-D},
number={9},
pages={1798-1802},
abstract={The GMM-UBM framework has been proved to be one of the most effective approaches to the automatic speaker verification (ASV) task in recent years. In this letter, we first propose an approximate decision function of traditional GMM-UBM, from which it is shown that the contribution to classification of each Gaussian component is equally important. However, research in speaker perception shows that a different speech sound unit defined by Gaussian component makes a different contribution to speaker verification. This motivates us to emphasize some sound units which have discriminability between speakers while de-emphasize the speech sound units which contain little information for speaker verification. Experiments on 2006 NIST SRE core task show that the proposed approach outperforms traditional GMM-UBM approach in classification accuracy.},
keywords={},
doi={10.1587/transinf.E92.D.1798},
ISSN={1745-1361},
month={September},}
Copy
TY - JOUR
TI - Approximate Decision Function and Optimization for GMM-UBM Based Speaker Verification
T2 - IEICE TRANSACTIONS on Information
SP - 1798
EP - 1802
AU - Xiang XIAO
AU - Xiang ZHANG
AU - Haipeng WANG
AU - Hongbin SUO
AU - Qingwei ZHAO
AU - Yonghong YAN
PY - 2009
DO - 10.1587/transinf.E92.D.1798
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E92-D
IS - 9
JA - IEICE TRANSACTIONS on Information
Y1 - September 2009
AB - The GMM-UBM framework has been proved to be one of the most effective approaches to the automatic speaker verification (ASV) task in recent years. In this letter, we first propose an approximate decision function of traditional GMM-UBM, from which it is shown that the contribution to classification of each Gaussian component is equally important. However, research in speaker perception shows that a different speech sound unit defined by Gaussian component makes a different contribution to speaker verification. This motivates us to emphasize some sound units which have discriminability between speakers while de-emphasize the speech sound units which contain little information for speaker verification. Experiments on 2006 NIST SRE core task show that the proposed approach outperforms traditional GMM-UBM approach in classification accuracy.
ER -