摘要
高斯混合模型(GMM)是进行说话人无关的语言辨识的一种有效方法,高斯混合二元模型(GMBM)是GMM模型的二元时序扩展,该文在GMBM和GMM-UBM模型的基础上提出了一种基于GMBM-UBBM模型的语言辨识系统,并利用OGI-TS电话语音库对算法的性能进行了测试,然后给出了实验结果。实验结果表明,该算法也是进行语言辨识的一种有效方法,与传统的GMM-UBM算法相比,该算法最多可以获得4.378%的相对改善率。
Gaussian Mixture Model is an effective method for speaker -independent language identification.Gaussian Mixture Bigram Model integrates bigram time correlation to extend the GMM.In this paper,a language identification algorithm using GMBM-UBBM is proposed based on GMBM and GMM-UBM,and some experiments are conducted using OGI-TS multi-language telephone speech corpus.Simulation results demonstrate the effectiveness of GMBM-UBBM for language identification tasks and use of this model allows the proposed system to distinguish among the three languages with maximal4.378%improvement accuracy superior to conventional GMM-UBM.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第3期29-32,共4页
Computer Engineering and Applications
基金
国家自然科学基金资助项目(批准号:60372038)