兰州方言的转换.pdfVIP

兰州方言的转换 摘要：本文着重研究并实现兰州方言单字的转换。语音转换技术是指改变源 说话人的语音特征使之具有目标说话人特征的一项技术。要实现转换的目标， 先要获得源说话人和目标说话人的语音特征。这部分工作就是前期的语音分析。 语音分析过程包括，以一定得原则设计和建立文本语料库，用Cool Edit Pro2.0软件录制语音语料，用VisualSpeech软件切分语音语料，最后是对切分好的 语音语料进行声学特征的分析。然后，在语音分析的基础上，建立模型实现转换 。我们运用普通话的归一化字调模型，分别对普通话和兰州话的四种声调的基频 进行参数化建模，然后，对参数化模型实现曲线拟合，得到一个四次曲 来表示 四种调型的基频曲 。最后用兰州话的基频曲 取代普通话的基频曲 ，最终重 新合成出的兰州方言。 关键字：语音特征基频归一化字调模型方言转换 straight 兰州方言的转换 ABSTRACT The thesis introduced a novel method for Lanzhou dialect conversion. Speech technology is a means to change the source of the speakers voice characteristics to the characteristics of the speaker has the goal of a technical. To achieve the goal of conversion, the first words to get the source and target. This work is part of the pre- speech analysis. Including voice analysis, Certain principles of the design and establishment of a text corpus, using the software Cool Edit Pro2.0 corpus voice recording, and voice Visual Speech corpus segmentation software, the final segmentation is a good acoustic characteristics of voice corpus analysis. Then, in the speech analysis based on the model to achieve the conversion. We use the normalized Mandarin word transfer model, Putonghua and Lanzhou, respectively, then the four tones of the fundamental frequency for parametric modeling, Then, the realization of the parameterized model curve fitting, a four times to indicate the four transfer curves of the fundamental frequency curve type. Finally, Lanzhou dialect, then to replace the fundamental frequency curve frequency curve of mandarin, and ultimately re-synthesis of the Lanzhou dialect t. Keywords: Speech features, Fundamental frequency , Normalized word transfer model Dialect conversion , straight - 1 - 兰州方言的转换