首页 | 官方网站   微博 | 高级检索  
     

一种新颖的混合语音检索算法*
引用本文:王让定,袁旭海,徐霁.一种新颖的混合语音检索算法*[J].计算机应用研究,2008,25(5):1349-1351.
作者姓名:王让定  袁旭海  徐霁
作者单位:宁波大学,纵横智能软件研究所,浙江,宁波,315211
基金项目:国家自然科学基金 , 浙江省宁波市科技攻关项目
摘    要:提出了一种新颖的混合语音检索算法。利用ICA的盲分离特性,将包含两个人声音的混合语音分离成只含一个人的单一语音,以分离出来的单一语音为新样本建立新的语音库,并与基于混合语音的语音库建立链接关系。查询时,利用DTW技术进行匹配检索,实现了基于混合语音的检索功能。实验表明该方法能够在混合语音中查询到用户所要求的记录。

关 键 词:混合语音  语音检索  独立分量分析  Mel频率倒谱系数  动态时间归整
文章编号:1001-3695(2008)05-1349-03
收稿时间:2008/4/20 0:00:00
修稿时间:2007年1月13日

Novel mixing speech retrieval algorithm
WANG Rang ding,YUAN Xu hai,XU Ji.Novel mixing speech retrieval algorithm[J].Application Research of Computers,2008,25(5):1349-1351.
Authors:WANG Rang ding  YUAN Xu hai  XU Ji
Affiliation:(CKC Software Lab, Ningbo University, Ningbo Zhejiang 315211, China)
Abstract:A novel mixing speech retrieval algorithm was presented.The characteristics of BBS(blind source separation) of ICA(independent component analysis) was used.The mixing speech records which contained a couple of different person's voice were separated into single speech records which contained only one person's voice.New speech database based on single speech records was established.The linkage between new speech database and original speech database which contained mixinged speech records was established.In query phrase,DTW(dynamic time warping) algorithm was used for retrieval matching and the function of retrieval based on mixing speech was achieved.Experimental result shows that algorithm of this paper can find demand records in mixing speech records.
Keywords:mixing speech  speech retrieval  ICA  MFCC  DTW
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《计算机应用研究》浏览原始摘要信息
点击此处可从《计算机应用研究》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号