计算机工程与应用 ›› 2010, Vol. 46 ›› Issue (19): 211-213.DOI: 10.3778/j.issn.1002-8331.2010.19.061

• 工程与应用 • 上一篇    下一篇

基于PCA技术的神经网络说话人识别研究

俞利强1,2,马道钧1,2   

  1. 1.西安电子科技大学通信工程学院,西安710071
    2.北京电子科技学院计算机科学与技术系,北京100070
  • 收稿日期:2008-10-06 修回日期:2010-03-01 出版日期:2010-07-01 发布日期:2010-07-01
  • 通讯作者: 俞利强

Neural network for speaker recognition of PCA technology

YU Li-qiang1,2,MA Dao-jun1,2   

  1. 1.College of Communication Engineering,Xidian University,Xi’an 710071,China
    2.Beijing Electronic Science and Technology Institute,Beijing 100070,China
  • Received:2008-10-06 Revised:2010-03-01 Online:2010-07-01 Published:2010-07-01
  • Contact: YU Li-qiang

摘要: 针对当提取以整段语音的多维语音特征参数为BP 神经网络输入而带来的说话人识别率和网络训练稳定性的问题,提出了一种用于BP 网络的基于主分量分析的PCA新方法。将该方法得到的降维语音特征参数用于BP 网络中,其识别率和训练速度都得到较大提高,使得基于BP 神经网络的说话人识别得到更好效果。

Abstract: For BP neural network on speaker recognition by extracting the entire multi-dimensional voice characteristic parameters
brought the recognition rate and the network stability problems.This paper proposes a new method of Principal
Components Analysis(PCA) on voice character parameters which used in BP.The new characters parameters can enhance the
recognition rate and training speed.This method can enhance the BP speaker recognition efficiency.

中图分类号: