作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2007, Vol. 33 ›› Issue (03): 35-36. doi: 10.3969/j.issn.1000-3428.2007.03.013

• 博士论文 • 上一篇    下一篇

基于子空间聚类的快速高斯计算

刘 斌1,谢凌云2   

  1. (1. 中国科学院声学研究所,北京 100080;2. 中国传媒大学传播声学研究所,北京 100024)
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-02-05 发布日期:2007-02-05

Fast Gaussian Computation Based on Subspace Clustering

LIU Bin1, XIE Lingyun2   

  1. (1. Institute of Acoustics, Chinese Academy of Sciences, Beijing 100080; 2. Institute of Communication Acoustics, Communication University of China, Beijing 100024)
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-02-05 Published:2007-02-05

摘要: 针对嵌入式语音识别系统,实现了基于子空间聚类的快速高斯计算,简化了HMM模型的计算复杂度,回避了声学模型重新训练的问题。在嵌入式系统上的实验数据表明,识别速度能获得20%以上的提高,而且识别率没有大幅降低。

关键词: 语音识别, 子空间聚类, 高斯计算, 嵌入式系统

Abstract: This paper presents a fast Gaussian computation algorithm based on subspace clustering of embedded speech recognition system to reduce the computation complexity. This algorithm does not need re-training model parameters. The experiments on embedded system show that the speed can be improved by above 20% and the recognition rate does not degrade much.

Key words: Speech recognition, Subspace clustering, Gaussian computation, Embedded system