首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于GA/VQ的说话人辨认的研究与实现
引用本文:王金明,李恩波.基于GA/VQ的说话人辨认的研究与实现[J].解放军理工大学学报,2005,6(3):214-218.
作者姓名:王金明  李恩波
作者单位:解放军理工大学,通信工程学院,江苏,南京,210007;解放军理工大学,通信工程学院,江苏,南京,210007
摘    要:为了改善在矢量量化说话人识别中,采用模板(码书)表征说话人,模板的质量对识别系统的性能。采用遗传算法改进模板的生成方式,构建了一种GA/VQ说话人辨认系统,给出了一种GA/VQ识别算法,通过遗传操作获得全局优化的说话人模板。实验证明,GA/VQ方法提高了码书的质量,比经典矢量量化识别系统识别率高。

关 键 词:矢量量化  遗传算法  说话人辨认  与文本无关  线性预测倒谱系数
文章编号:1009-3443(2005)03-0214-05
修稿时间:2004年11月30

Research and implementation of speaker identification based on GA /VQ
WANG Jin-ming and LI En-bo.Research and implementation of speaker identification based on GA /VQ[J].Journal of PLA University of Science and Technology(Natural Science Edition),2005,6(3):214-218.
Authors:WANG Jin-ming and LI En-bo
Institution:Institute of Communications Engineering, PLA Univ. of Sci. & Tech.,Nanjing 2100 07, China;Institute of Communications Engineering, PLA Univ. of Sci. & Tech.,Nanjing 2100 07, China
Abstract:In speech signal processing, the vector qua ntization(VQ) is an important method. The speaker recognition method based on VQ is researched in this paper. In VQ speaker recognition, codebook was adopted t o represent the speaker, and the performance of system is mainly influenced by the qua lity of codebook. Genetic algorithm was adopted in this paper to get global opti m ized speaker codebook and a GA/VQ speaker identification algorithm was proposed. E xperiment results show that this method improves the quality of codebook, the sy s tem based on GA/VQ performs higher correct identification rates compared with th at based on conventional VQ method.
Keywords:VQ(vector quantization)  GA(genetic algorithm)  speaker identification  text-in dependent  LPCC (linear prediction cepstrum coefficients)
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《解放军理工大学学报》浏览原始摘要信息
点击此处可从《解放军理工大学学报》下载免费的PDF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