基于古音系统的汉语方言语音合成

Speech synthesis for Chinese dialects based on the phonological system of ancient Chinese

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。

The HTS speech training synthesis tools and STRAIGHT speech synthesizer are used for speech training synthesis of an unknown dialect phonetic system.This study applies these methods to various Chinese dialects by designing a set of universal pronunciation texts for Chinese dialect speech synthesis with recording of a synthesized speech corpus for some Chinese dialects.A Chinese dialect speech synthesis platform is then built based on the phonological system of ancient Chinese.Results for Mandarin speech synthesis show that the synthesized sound based on the phonological system of ancient Chinese is quite close in terms of intelligibility and quality to the synthesized sound based on the pinyin system.Thus,the speech synthesis based on the phonological system of ancient Chinese for Chinese dialects is effective and feasible.