语言辨识的矢量量化方法(VQ)

Vector Quantization Used in Language Identification

与语音处理的其他领域相比 ,自动语言辨识是一个相对较新的研究 ,而且是一项较难的课题。本文给出了语言辨识的两种矢量量化方法 ,并利用OGI TS电话语音库对 2种算法的性能进行了测试 ,然后给出了实验结果。 2种算法的最佳识别率分别为 6 5 .3 8%和 73 .0 8%。

Compared with speech technologies in speech processing, automatic language identification is a relatively new yet difficult problem. In this paper,two language identification algorithms are provided and some experiments are conducted using OGI-TS telephone speech corpus. Then experiments results are described.The optimal recognition rates for the two algorithms are 65.38% and 73.08%, respectively.