期刊文献+

基于可区分性加权的模糊核说话人识别 被引量:2

A Fuzzy Kernel with Discriminative Weighted Method for Speaker Recognition
下载PDF
导出
摘要 针对训练和识别语音数据较少的情况,本文提出了一种新的说话人识别算法.通过核映射,在高维特征空间对说话人的语音特征进行模糊矢量量化.为了增加说话人之间的可区分性,提出了一种基于高维特征空间的码字矢量的权值分配方法,对具有较强区分性的码字矢量分配较大的权值,并将产生的权值和说话人的码书一起形成说话人数据库.识别时,提出一种模糊核加权最近邻近分类器,在高维特征空间中对说话人进行匹配.实验表明,该算法在训练语音少于8s,识别语音为1s时,能够得到较好的识别结果. As to small amounts of training and test speech data, it proposed a new speaker recognition algorithm. By the kernel mapping,it used the fuzzy vector quantization to quantize the speakers' speech features in the high dimensional feature space.in order to improve the discriminations of different speakers, it presented a novel weights assignment method. It assigned the lager weight to the code vector with higher discriminative power. Then, it used the codebooks and weights to form the speakers' database. In the matching phase, it proposed a fuzzy kernel weighted nearest prototype classifier, which can identify different speakers in the high dimensional space.Experimental results show that when the Iraining speech data is less than 8s,and test speech data is 1 s, this algorithm can get good performance.
出处 《电子学报》 EI CAS CSCD 北大核心 2008年第7期1446-1450,共5页 Acta Electronica Sinica
关键词 说话人识别 少量语音数据 可区分性权值 模糊核加权最近邻近分类器 模糊核矢量量化 speaker recognition small amounts of speech data discriminative weighted value fuzzy kernel weighted nearest prototype classifier fuzzy kernel vector qnantization.
  • 相关文献

参考文献13

二级参考文献31

  • 1刘静,钟伟才,刘芳,焦李成.免疫进化聚类算法[J].电子学报,2001,29(z1):1868-1872. 被引量:43
  • 2刘健庄,谢维信,黄建军,李文化.聚类分析的遗传算法方法[J].电子学报,1995,23(11):81-83. 被引量:27
  • 3朱民维.计算机语音技术[M].北京:北京航空航天大学出版社,1991.39-86.
  • 4马卡尔 娄乃英(译).语音信号线性预测(第一章)[M].北京:中国铁道出版社,1997..
  • 5刘增良.模糊技术与神经网络技术选编[M].北京:北京航天航空大学出版社,1995.120-157.
  • 6F Soong et al.A vector quantization approach to speaker recognition[J]. IEEE ICASSP, 1985:387-390.
  • 7K S Asultan,S Selim.A global algorithm for the fuzzy clustering problem[J].Pattem Recognition, 1993 ;26(9) : 1357-1361.
  • 8CAMPBELL J P. Speaker recognition: a tutorial[J]. Proceedings of the IEEE, 1997,85(9): 1437-1462.
  • 9PETRY A, BARONE D A C. Speaker identification using nonlinear dynamical feature[J]. Chaos, Solitons & Fractals, 2002, 13(2):221-231.
  • 10WAN V, CAMPBELL W M. Support vector machines for speaker verification and identification[A]. Proceedings of the 2000 IEEE Signal Processing Society Workshop[C]. 2000. 775-784.

共引文献30

同被引文献16

  • 1安冬,王守觉.基于仿生模式识别和PCA/ICA的DOA估计方法[J].电子学报,2004,32(9):1448-1451. 被引量:14
  • 2张庆芳,赵鹤鸣.基于改进VQ算法的文本无关的说话人识别[J].计算机工程与应用,2006,42(10):65-68. 被引量:7
  • 3P Joseph, JR Campbell. Speaker recognition: A tutorial[ J]. Proceedings of the IEEF,, 1997,85(9) : 1437 - 1462.
  • 4Tomi Kinntmen, Li Haizhou. An overview of text-independent speaker recognition: From feature to super vectors [J]. Speech Communication, 2009,52(2) : 12 - 40.
  • 5I T Jolliffe. Principal Component Analysis [ M ]. Springer: Berlin, 1986.
  • 6N Kambhatla. Dimension reduction by local PCA [J]. Neural Computing, 1997,9 (7) : 1493 - 1516.
  • 7C W Seo, K Y lee. GMM based on local PCA for speaker identification [ J ]. Electronics Letters, 2001,37 ( 24 ) : 1486 - 1488.
  • 8S Molau, M Pitz, R Schluter. Computing Mel-frequency cepstral coefficients on the power spectrum [ A ]. Proceedings of the 2001 IEEE International Conference on Acoustics, Speech and Signal Processing[ C]. USA: IEEE Press, 2001.73 - 76.
  • 9S Furui. Cepstral analysis technique for automatic speaker verification[ J]. IEEE Transactions on Acoustics, Speech and Signal Processing, 1981,29 (2) : 254 - 271.
  • 10J C Bezdek. Pattem Recognition with Fuzzy Objective Function Algorithm[ M]. New York: Plenum Press, 1981.

引证文献2

二级引证文献9

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部