基于GMM的说话人识别，搭建了一个说话人识别系统用于试验测试，验证了一些参数对性能

的影响，同时使用了多线程并行处理技术，以此缩短识别时间:并提出了一种放

大特征向量差距，变换特征向量在特征空间的分布来提升大容量语音库中说话人

识别率的方法。

-GMM-based speaker recognition, to build a speaker recognition system used for pilot testing to verify the performance impact of some of the parameters, using a multi-threaded parallel processing technology, in order to shorten the recognition time: and presents a zoom feature Vector gap, transformed feature vector in the feature space to improve the distribution of large speech database speaker recognition rate method.