New score normalization method in speaker verification

Journal of Computer Applications

• Typical applications • Previous Articles Next Articles

New score normalization method in speaker verification

<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>X<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>i<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>-<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>J<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>i<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>a<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>G<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>a<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>o<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>D<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>a<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>n<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>Q<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>u<a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a> <a href="http://www.joca.cn/EN/article/advancedSearchResult.do?searchSQL=((([Author]) AND 1[Journal]) AND year[Order])" target="_blank"></a>

Received:2007-05-16 Revised:1900-01-01 Online:2007-10-01 Published:2007-10-01
Contact: Xin-Jian Gao

说话人确认中分数归一化的一种新方法

高新建屈丹李弼程

河南河南

通讯作者: 高新建

Abstract

Abstract: In the speaker verification field, the distributions of the target and imposter scores are bimodal, and the output score distributions of different target speaker models are different too, so it is difficult to estimate a common threshold. The most frequently used score normalization techniques: T-Norm and Z-Norm were analyzed. A new technique D-Norm was introduced based on the KL distance. Then a new method ZD-Norm was proposed based on Z-Norm and D-Norm. The experiments show that the ZD-Norm method can improve the performance efficiently and make the estimation of threshold easier.

Key words: speaker verification, score normalization, Z-Norm, D-Norm

摘要： 在说话人确认中，由于目标说话人和冒认者的得分分布是双峰分布，并且不同目标说话人模型得分分布不一致，使对所有说话人确定一个统一的阈值变得困难，导致系统性能下降。分数归一化通过调整冒认者的得分分布来调整阈值。简要介绍了目前最常用的两种归一化方法：零归一化（Z-Norm）和测试归一化(T-Norm)。重点引入了一种新的根据KL距离的D-Norm 归一化方法。然后结合Z-Norm 和D-Norm的优点，又提出一种新的方法ZD-Norm。对这四种归一化方法的性能进行了比较。实验表明，ZD-Norm相对Z-Norm和D-Norm，能够更有效地提高说话人确认系统的性能。

关键词: 说话人确认, 分数归一化, Z-Norm, D-Norm

Xin-Jian Gao Dan Qu . New score normalization method in speaker verification[J]. Journal of Computer Applications.

高新建屈丹李弼程 . 说话人确认中分数归一化的一种新方法[J]. 计算机应用.

References

Related Articles 4

Recommended Articles

Metrics