基于因特网的说话人识别技术研究  

The Study on Speaker Recognition Technology Based on Internet

作　　者：叶蕾[1] 方鹏[2] 杨震[1] 

机构地区：[1]南京邮电学院信息工程系,江苏南京210003 [2]南京邮电学院通信工程系,江苏南京210003

出　　处：《南京邮电学院学报（自然科学版）》2004年第3期45-48,52,共5页Journal of Nanjing University of Posts and Telecommunications

摘　　要：比较了基于因特网的说话人识别技术中Mel倒谱特征各阶参数的抗噪性能,并分析和验证了交织及丢失数据替代技术对改善基于因特网的说话人识别系统性能的重要作用。In this paper, we compare the noise-resistance characteristic of each of Mel cepstral coefficients, which are applied to speaker recognition on Internet. We also analyze and prove the advantages of interleaving and missing packet substitution technologies for the improvement of the performance of the speaker recognition system.

关 键 词：因特网 说话人识别 分组丢失 交织 Mel倒谱参数 

分 类 号：TN912.34[电子电信—通信与信息系统] TP393.4[电子电信—信息与通信工程]