论文部分内容阅读
采用HTS语音训练合成工具和STRAIGHT语音合成器,尝试在未知某方言(或土语)的实际语音系统的条件下开展相应的语音训练合成工作。采用古音系统来转写汉语字音,并在此基础上设计相应的问题集以实现语音训练合成。设计了一套面向汉语方言语音合成的通用发音文本;录制了一些汉语方言点的合成语音语料库;搭建了基于古音系统的汉语方言语音合成平台。普通话的合成实验结果表明:基于古音系统训练合成出来的语音,在可懂度和音质上跟基于普通话拼音系统训练合成出来的语音非常接近。这表明基于古音系统进行汉语方言语音合成的方法是有效的、可行的。
Using HTS speech training synthesis tool and STRAIGHT speech synthesizer, we try to synthesize speech training without knowledge of the actual speech system of a dialect (or Turkic language). Using ancient sound system to transfer Chinese pronunciation, and on the basis of the design of the corresponding problem set to achieve voice training and synthesis. A set of generalized phonetic texts for speech synthesis of Chinese dialects was designed. Some synthetic phonetic corpora of Chinese dialects were recorded. A Chinese dialect speech synthesis platform based on the ancient phonetic system was constructed. The results of mandarin synthesis experiment show that the speech synthesized by the training of ancient phonetic system is very close to the synthesized phonetic system based on Mandarin pinyin system in intelligibility and sound quality. This shows that the method of Chinese dialect speech synthesis based on ancient sound system is effective and feasible.