Abstract: This paper proposes a new method, which uses MFCC combined with complexity. Some experiments are conducted by using OGI-TS telephone speech corpus. Vector quantization method is employed to recognize three languages (English, Mandarin and Japanese), and it analyzes the effects of language identification. It is shown that the method can remarkably improve the recognition accuracy and robustness.

Key words: language identification, complexity, standard vector quantization

摘要： 提出一种在传统提取MFCC特征的基础上增加复杂性特征的方法，利用OGI-TS电话语音库对该方法进行性能测试，比较、分析英语、汉语、日语3个语种的识别效果，结果表明，该方法相对于传统方法能明显提高语种识别的准确性和鲁棒性。

关键词: 语种辨识, 复杂性, 标准矢量量化

CLC Number: 

• TP18