摘要
介绍了一个基于GMM实时说话人识别系统的设计与实现,系统具有实时说话人辨认和实时说话人确认功能。在实验室条件下,对不同的高斯混合密度个数及采样率进行了测试,测试了模型的自适应性能。实验表明系统具有较好的识别准确率。
The design and implementation of a real-time speaker recognition system which is based on GMM （Gaussian Mixture Model） are presented. The system has the characteristics of real time speaker identification and real time speaker verification. In the lab environment, the performance of the system, as well as the model adaptation, has been fully tested with GMMs of different numbers of Gaussian mixtures and different sampling rates. The testing results show that the GMM-based system has a satisfactory correctness in performing speaker recognition.
出处
《电声技术》
2007年第6期53-55,共3页
Audio Engineering
基金
厦门大学"985工程"二期"信息技术"创新平台项目资助
项目编号0000-X07204
