Automatic voice building for flexible speech synthesis

随着大型语音库的日益普及，通过应用统计学习算法来构建语音合成系统成为可能，这种方法被称为数据驱动或基于语料库的方法。这些系统可以自动训练，不仅可以生成自然、高质量的合成语音，而且可以再现原始说话人的语音特征。然而，为了使整个语音构建过程完全自动化，我们需要以自动的方式构建语音库。在这项研究工作中，我们研究了一个基于HMM的语音合成系统的自动语音生成技术，该系统可以合成不同音质的语音。首先，我们实现了一个基于图形用户界面的标注工具，称为PLED(韵律和语言标签编辑器)。然后，为了构建一个自动语音生成系统，我们开发了一种自动口音标注技术。实际应用表明，该系统成功地实现了口音信息的标注。

The increasing availability of large speech databases makes it possible to construct speech synthesis systems, which are referred to as data-driven or corpus-based approach, by applying statistical learning algorithms. These systems, which can be automatically trained, not only generate natural and high quality synthetic speech but also can reproduce voice characteristics of the original speaker. However, to make the whole voice building process fully-automatic, we need to construct speech databases in an automatic way. In this research work, we investigate automatic voice building techniques for an HMM-based speech synthesis system which can synthesize speech with various voice qualities. First, we implemented an GUI-based labeling tool, called PLEd (Prosody and Linguistic Label Editor). Then, in order to construct an automatic voice building system, we have developed an automatic accent labeling technique. It has been shown that by using the developed system, we have successfully label accent information.