基于新的MRSVM说话人辨识方法

Speaker identification method based on novel MRSVM

提出一个新的基于MRSVM的说话人辨识方法,首先对语音特征矢量进行LDA降维,得到具有区分力的特征矢量,然后对其进行模糊核聚类,根据样本选择算法,选择聚类边界的特征矢量作为支持向量训练支持向量机,在不影响识别率的情况下,大大减少了支持向量机的存储量和训练量。实验表明该方法具有较好的总体效果。

A speaker identification method is proposed based on a novel Multi-Reduced Support Vector Machine（MRSVM）.Firstly,speech feature dimensions are reduced by using LDA transform;Secondly,the training data are selected at boundary of each cluster as Support Vectors（SVs） by using kernel-based fuzzy clustering technique.The experiment results show that the training data,time and storage can be reduced remarkably by using the proposed method without deteriorating recognition performance.The method is proved to be effective by the experiments.