Abstract: This paper designs Web authentication system based on voiceprint recognition. An ActiveX control is developed in the browser to record and encapsulate the voice properties of the users and then transfer to the server. In the server side, a Hidden Markov Model(HMM) is designed to represent the Chinese words which are separated by an inserted silence state. And then a stable voiceprint database is constructed after the enough training of speech samples. Actual test results show that this system is suitable for controlling Web resource and has good identification performance and strong portability.

Key words: identity authentication, voiceprint recognition, Hidden Markov Model(HMM), Dynamic Time Warping(DTW)

摘要： 设计一个基于声纹识别的Web身份认证系统。在浏览器端利用自主开发的ActiveX录音控件录取封装使用者的声纹特征并传送给服务器，服务器端使用隐马尔科夫模型表述单字，单字之间通过增加静音状态分割以及语音训练形成稳定的声纹特征模型作为身份验证的基础库。实际测试结果表明，该系统识别性能好、可移植性强，适用于网络资源的远程声控。

关键词: 身份认证, 声纹识别, 隐马尔科夫模型, 动态时间校正

CLC Number: 

• TP309