计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (29): 225-227.

• 工程与应用 • 上一篇    下一篇

基于贝叶斯算法与高斯混和模型的语者确认研究

胡海波1,傅 鹂1,向 宏1,周 元2, 刘晓艳2   

  1. 1.重庆大学 软件学院,重庆 400044
    2.重庆大学 数理学院 重庆 400044
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-10-11 发布日期:2007-10-11
  • 通讯作者: 胡海波

Speaker verification based on Bayesian adaptation and Gaussian Mixture Model

HU Hai-bo1,FU Li1,XIANG Hong1,ZHOU Yuan2,LIU Xiao-yan2   

  1. 1.College of Software Engineering,Chongqing University,Chongqing 400044,China
    2.College of Mathematics and Physics,Chongqing University Chongqing 400044,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-10-11 Published:2007-10-11
  • Contact: HU Hai-bo

摘要: 文章针对统一背景模型与群模型两种反模型进行了分析,在基于统一背景模型与群模型的改进说话人确认模型的基础上,将贝叶斯自适应算法引入到基于高斯混合统一背景模型的说话人确认系统,解决了说话人确认中存在的模型不匹配问题,通过文本无关的测试语音库进行的实验和分析显示,改进算法具有更好的识别效果。

Abstract: In this paper,Universal Background Model and Cohort Model were combined to improve the performance of Gaussian Mixture Model for Speaker Verification.And Bayesian maximum a posteriori estimation has been used for training a speaker model from background model,to solve the problem of model miss matching in speaker verification system.Experiments have been based on text-independent speech corpus.The result shows that the approach above has better performances than the origin Gaussian Mixture Model.