摘要
高斯混合模型( GMM)是当今说话人识别的一种流行算法,但 GMM的训练的目标是使似然度最大,并不能产生识别性能最佳的模型。本文提出了GMM +MCE(最小分类错误)的模型来解决这一问题。并通过实验证明了其有效性。
Gauss Mixture Model(GMM) is a popular algorithm of speaker recognition. However the object of training GMM is to maximize the likelyhood, which can limit the performance of the recognition system. In this paper the model of GMM+MCE (Minimum Classification Error) is proposed to solve this problem and the effectiveness of the algorithm is verified by experiments.
出处
《电路与系统学报》
CSCD
2000年第3期46-49,共4页
Journal of Circuits and Systems
关键词
高斯混合模型
最小分类错误
说话人识别
Gauss mixture model
minimum classification Error
speaker Recognition