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.
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
关键词
因特网
说话人识别
分组丢失
交织
Mel倒谱参数
Internet
Speaker Recognition
Mel cepstral coefficients
Interleaving
Packet-Loss
