论文部分内容阅读
在HMM的基础上,针对少量的训练样本情况,提供了一种新的训练算法—MCE(Minimum Classification Error)算法,并利用该算法建立了一个局部范围内不依赖于文本的说话人辨认系统,当每个说话人的样本训练量为10时,系统识别的正确率达到了97.14%。
On the basis of HMM, a new training algorithm, MCE (Minimum Classification Error) algorithm, is provided for a small number of training samples. A speaker-independent speaker recognition system is established by using this algorithm. When the sample training amount per speaker is 10, the correct rate of system identification reaches 97.14%.