Abstract:
This paper sets up a new Support Vector Machine-Gussian Mixture Model(SVM-GMM) to improve speaker recognition rate based on open-set. The basic idea of the new model is that the classification results of the SVM are confirmed with GMM. Due to the good classification performance of SVM and the good description of the internal similarity of some category of GMM, the good recognition effect can be obtained by combining the two models. Experimental results show that using SVM-GMM model can improve the open-set speaker recognition rate effectively.
Key words:
Support Vector Machine(SVM),
Gussian Mixture Model(GMM),
open-set speaker recognition,
Equal Error
Rate(EER)
摘要: 建立一种支持向量机-高斯混合模型(SVM-GMM),用以提高开集说话人识别的识别率。该模型的基本思想是将SVM的分类结果用GMM模型进行确认。由于SVM模型具有较好的分类性能,而GMM模型能够较好地描述类别内部的相似性,因此这2个模型的组合能够优势互补,从而获得较好的识别效果。实验结果表明,使用SVM-GMM模型能有效地提高开集说话人识别的识别率。
关键词:
支持向量机,
高斯混合模型,
开集说话人识别,
等误识率
CLC Number:
CHEN Li, XU Dong-Beng. Method of Open-set Speaker Recognition Based on SVM-GMM[J]. Computer Engineering, 2011, 37(14): 172-174.
陈黎, 徐东平. 基于SVM-GMM的开集说话人识别方法[J]. 计算机工程, 2011, 37(14): 172-174.