Articulatory-feature based sequence kernel for high-level speaker verification
Articulatory-feature based sequence kernel for high-level speaker verification
复制标题
DOI：
10.1109/icmlc.2008.4620884
复制
发表时间：
2008-07
期刊：
影响因子：
--
通讯作者：
Shi-Xiong Zhang;M. Mak
中科院分区：
文献类型：
--
作者：
Shi-Xiong Zhang;M. Mak
研究表明，基于发音特征的语音类发音模型（AFCPM）可以捕获说话者的发音特征。然而，AFCPM 中使用的评分方法并未明确使用训练数据中可用的判别信息。为了利用这些信息，本文提出通过堆叠 AFCPM 中的离散密度将说话人模型转换为超向量。 AF 内核是由目标说话者、背景说话者和声明者的超向量构建的。然后训练基于 AF 内核的 SVM 对超向量进行分类。结果表明，AF 核评分与似然比评分是互补的，当两种评分方法结合使用时，可以获得更好的性能。
Research has shown that articulatory feature-based phonetic-class pronunciation models (AFCPMs) can capture the pronunciation characteristics of speakers. However, the scoring method used in AFCPMs does not explicitly use the discriminative information available in the training data. To harness this information, this paper proposes converting speaker models to supervectors by stacking the discrete densities in AFCPMs. An AF-kernel is constructed from the supervectors of target speakers, background speakers, and claimants. An AF-kernel based SVM is then trained to classify the super-vectors. Results show that AF-kernel scoring is complementary to likelihood-ratio scoring, leading to better performance when the two scoring methods are combined.