基于MVQM说话人识别系统的DSP实现  

Realizing Speaker Recognition System Based on MVQM with DSP

作　　者：谢建平[1] 

机构地区：[1]丽水学院计算机与信息工程学院,浙江丽水323000

出　　处：《微电子学与计算机》2006年第6期220-222,226,共4页Microelectronics & Computer

摘　　要：文章运用TMS320C5416实现了说话人自动识别系统。提出了一种新的说话人识别方法。该方法综合了VQ和GMM的优点。通过用VQ误差尺度取代传统GMM的输出概率函数,减少了建模时对训练数据量的要求,提高了识别速度。实验结果证明该方法是有效的。In this paper, TMS320C541 was used to realize auto speaker recognition system. For the integration of merits of VQ and GMM, a new method of speaker recognition was presented. When VQ error scale was adopted instead of the probability output by tradition GMM, data capacity request was reduced during modeling and reached a high rate recognition. The result also showed that the method is effective to speaker recognition.

关 键 词：说话人识别 矢量量化 混合高斯模型 数字信号处理器 

分 类 号：TN912.34[电子电信—通信与信息系统]