Computer Engineering and Applications ›› 2013, Vol. 49 ›› Issue (2): 174-176.

Previous Articles     Next Articles

GSV-SVM based language recognition system

LIANG Chunyan1, AN Maobo2, LIU Zhenye2, SUO Hongbin1, WANG Junjie1   

  1. 1.Key Lab of Speech Acoustics and Content Understanding, Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China
    2.National Computer Network Emergency Response Technical Team/Coordination Center of China, Beijing 100029, China
  • Online:2013-01-15 Published:2013-01-16

高斯超向量-支持向量机鉴别性语种识别系统

梁春燕1,安茂波2,刘振业2,索宏彬1, 汪俊杰1   

  1. 1.中国科学院 声学研究所 语言声学与内容理解重点实验室,北京 100190
    2.国家计算机网络应急技术处理协调中心,北京 100029

Abstract: The Support Vector Machine(SVM) has been widely used in language recognition. It has reached comparable performance with the traditional Gaussian Mixture Model(GMM). Gaussian Super Vector-Support Vector Machine(GSV-SVM), which effectively combines GMM and SVM, is presented in this paper. The experiments are carried out on the NIST LRE2003 and LRE2007 test corpus. The results indicate that the GSV-SVM system can achieve significant improvements on the long-time duration test set, compared with the method of GMM.

Key words: language recognition, Gaussian mixture model, support vector machine, Gaussian super vector

摘要: 支持向量机在语种识别技术中获得了广泛的研究和应用,并且达到和传统混合高斯模型相当的性能。高斯超向量-支持向量机系统将高斯混合模型与支持向量机有效地结合起来,采用高斯超向量核函数,以支持向量机作为后端分类器。重点介绍基于高斯超向量-支持向量机的语种识别系统,并和传统的高斯混合模型系统进行比较。在美国国家标准技术研究院2003年和2007年语种识别评测数据集上进行实验。实验结果表明,高斯超向量-支持向量机系统相对于混合高斯模型建模的方法,在长时数据上有较明显的性能优势。

关键词: 语种识别, 高斯混合模型, 支持向量机, 高斯超矢量