基于特征均值距离的短语音段说话人聚类算法  

Feature Mean Distance Based Speaker Clustering for Short Speech Segments

作　　者：李艳雄[1] 吴永[1] 贺前华[1] 

机构地区：[1]华南理工大学电子与信息学院,广州510640

出　　处：《电子与信息学报》2012年第6期1404-1407,共4页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61101160;60972132);中央高校基本科研业务费专项基金(2011ZM0029);广东省自然科学基金博士启动项目(10451064101004651)资助课题

摘　　要：该文提出一种基于特征均值距离的短语音段说话人聚类算法。首先,定义特征均值距离用来在特征层而不是模型层刻画两个类之间的相似度;然后,迭代合并特征均值距离最小的两个类,直到任意两类之间的特征均值距离的最小值大于一个自适应门限为止。采用取自两个语音数据库的短于3 s的语音段进行实验测试,结果表明:与基于AHC+BIC的算法相比,F度量值平均提高了5%,运算速度约为以前算法的4.68倍。An algorithm of speaker clustering is proposed based on Feature Mean Distance（FMD） for short speech segments.First,a distance measure,i.e.FMD,is introduced to represent the similarities between two clusters on the level of feature instead of the level of model.Then,two clusters with the minimum of FMDs are iteratively merged until the minimum of FMDs is larger than an adaptive threshold.Experimental results show average 5% improvements in F measure are obtained in comparison with the AHC＋BIC based algorithm.In addition,the proposed algorithm is 4.68 times faster than the AHC＋BIC based algorithm.

关 键 词：语音信号处理 说话人聚类 特征均值距离 短语音段 

分 类 号：TN912.3[电子电信—通信与信息系统]