基于LLE和模糊核聚类的语音可视化仿真-JournalofNortheastern.PDFVIP

第30 卷第6期 东 北 大 学 学 报 ( 自 然 科 学 版 ) Vol30, No. 6 2 00 9 年 6 月 Journal of Northeastern U ni ersity( Natural Science) Jun. 2 0 0 9 LLE 韩志艳, 王 旭, 王 健, 薛丽芳 ( , 110004) : , ( LLE) LLE , , Mercer , , , , 10 10 720 ( ) , : ; ; ; ; : N 945. 13 : A : 1005-3026( 2009) 06-0790-04 Speech Visualization Simulation Based on LLE and Fuzzy Kernel Clustering Algorithm HA N Zhi-y an, WA N G X u, WA N G J ian, X UE Li-f ang ( School of Information Science & Engineering , Northeastern Uni ersity, Shenyang 110004, China. Correspondent: HAN Zhi-yan, E-mail: hanzyme @ 126. com) Abstract: According to the time- arying speech signal, a no el method combining LLE( locally linear embedding) w ith fuzzy kernel clustering algorithm w as proposed for speech isualization, w here LLE could reduce the nonlinear dimensionality of the speech features and then the fuzzy kernel clustering algorithm w as used for clustering analysis, i. e. the M ercer kernel function was used to change the data in original space into a high-dimensional eigenspace through nonlinear mapping, and then the fuzzy clustering analysis was made in the high-dimensional eigenspace. T hus, after the kernel function mapping, the original inherent features of speech w ere highlighted to impro e the position-based speech isualization. 720 data in Chinese ow els were obtained from 10 male and 10 female students speech in lab, the results of simulation experiments show the feasibility and alidity of the method. Key words: speech signal; isualization; LLE; kerne-l based method; fuzzy kernel clustering [ 1]