摘要
在基于智能卡的说话人确认系统中,实现了一种新的端点检测方法:能频值端点检测方法,其取得了较好的效果;在鲁棒性方面,研究了Mel倒谱系数各分量在说话人识别中的贡献,以及在参数级上Mel倒谱系数的差分系数及倒谱均值相减法对说话人识别的贡献;最后,讨论了基于智能卡的生物特征识别技术的应用途径。
In the smart card-based speaker verification system, energy frequency value endpoint detection is implemented. In the aspect of robustness, the contribution of each dimension of MFCC is studied, and the contribution of the differential coefficent of Mel-fequency cepstrum coefficient and cepstral mean subtraction are also discussed. Finally, the application of technology based on smart card and biometric verification is analyzed.
出处
《电声技术》
2006年第8期50-53,56,共5页
Audio Engineering
关键词
说话人确认
智能卡
MEL倒谱系数
倒谱均值相减法
能频值
speaker verification
smart card
Mel-frequency cepstrum coefficent
cepstral mean subtraction
energy frequency value
