新颖检测法在说话人识别技术中的应用

Application of Novel Detection Method in Speaker Recognition

提出一种用于说话人识别中说话人语音特征向量聚类的方法——新颖检测法。通过提取出的特征参数(MFCC和LPCC),建立系统模型,实验结果表明,将新颖检测法结合VQ用于特征向量的分类,较之于单纯的VQ分类,取得了识别率高、稳健型强、确认可靠的效果。

Proposes a novel detection method that applied for the clustering of feature parameters in speaker recognition. Speaker models are established with the feature parameters （MFCC ＆ LPCC）extracted from speakers, and the results indicate that combining novel detection ＇with VQ can get higher recognition rate than only with VQ, also the system is more stable and reliable.