摘要
源-目标说话人声音转换是一种变换说话人声音特征的技术,它将源说话人的声音转换成目标说话人的声音。本文选择声道共振峰参数作为待转换的特征参数,为了克服线性多变量回归转换方法(LMR)中分类不准带来的误差,采用基于径向基函数神经网络的非线性转换方法(RBFNN)获取转换规则。以5个普通话元音为实验,验证了分类数目和训练集对2种转换方法的影响。实验结果表明,RBFNN方法的转换效果优于LMR方法;并在只有较少训练集数据时也能得到较好的转换效果。
Voice conversion is a method which transforms the source speech to a speech signal with the acoustic characteristics of target speaker. Formant parameters which estimated by root-finding method based on LP analysis are chosen for the transformation parameters. A nonlinear transformation based on radial basis function neural network is presented to reduce transformation error caused by inaccurate classification of linear multivariate regression. Five vowel phones in Mandarin speech are selected and some experiments about the number of class and the training data are carried out. Experimental results prove that RBF neural network has a better performance than LMR and the performance of RBF neural network has litter relation with training data.
出处
《电子测量技术》
2006年第6期60-63,共4页
Electronic Measurement Technology
