应用科学学报 ›› 2010, Vol. 28 ›› Issue (1): 38-43.

• 信号与信息处理 • 上一篇    下一篇

嵌入自联想神经网络的高斯混合背景模型说话人确认

陈存宝, 赵力   

  1. 东南大学信息科学与工程学院,南京210096
  • 收稿日期:2009-03-16 修回日期:2009-10-20 出版日期:2010-01-20 发布日期:2010-01-20
  • 作者简介:陈存宝,博士生,研究方向:语音信号处理, E-mail: chencunbao@gmail.com;赵力,教授,博导,研究方向:语音信号处理,E-mail: zhaoli@seu.edu.cn
  • 基金资助:

    国家自然科学基金(No.60872073, No.60975017); 江苏省自然科学基金(No.BK2008291)资助

Speaker Verification Based on GMM-UBM with Embedded Auto-associate Neural Network

CHEN Cun-bao, ZHAO Li   

  1. School of Information Science and Engineering, Southeast University, Nanjing 210096, China
  • Received:2009-03-16 Revised:2009-10-20 Online:2010-01-20 Published:2010-01-20

摘要:

 提出在高斯混合背景模型中嵌入自联想神经网络的方法,并将它用于说话人确认. 该方法利用神经网络
和高斯混合背景模型各自的优点,以极大似然概率为训练准则,将两者作为一个整体进行训练,揭示了特征向量的
空间信息. 嵌入的神经网络起到了数据整形的作用,增强了目标说话人数据的相似性. 在背景模型和目标模型的训
练中交替更新高斯混合模型和神经网络的参数. 实验表明,采用本文提出的模型并结合TNorm方法,比基线系统
的确认率提高26%.

关键词: 说话人确认, 高斯混合背景模型, 自联想神经网络, 嵌入

Abstract:

This paper proposes to embed an auto-associate neural network (AANN) in the Gaussian mixed
model-universal background model (GMM-UBM) for speaker verification. The scheme integrates the merits of
both GMM and AANN. GMM and AANN are trained as a whole in terms of maximum likelihood (ML), and
spatial information of the feature vectors is disclosed. AANN reshapes distribution of the data and improves
similarity of intra-class data. In the training, parameters of GMM and AANN are updated alternately. Experimental
results show that the proposed method together with TNorm can provide improvement of verification
rate by 26% over the baseline GMM-UBM.

Key words: speaker verification, Gaussian mixed model-universal background model (GMM-UBM), autoassociate, neural network (AANN), embed

中图分类号: