基于可变长音素序列拼接单元的维吾尔语语音合成技术研究  

Uighur Speech Synthesis Technology Research Using Variable Length Phoneme Sequences as Concatenation Unit

作　　者：周艳[1] 艾斯卡尔[1] 

机构地区：[1]新疆大学信息科学与工程学院,乌鲁木齐830046

出　　处：《四川理工学院学报（自然科学版）》2007年第2期64-68,共5页Journal of Sichuan University of Science & Engineering(Natural Science Edition)

摘　　要：文章采用了一种以可变长音素序列为拼接单元的维吾尔语语音合成系统的技术方案,阐述了维吾尔语的语言特点及语音合成中必须考虑的语音协同发音等现象,给出了语音库的设计思路及其句子、短语、词语、音节以及音素等多级语音库结构,以便直接从语音库中找到拼接单元,还考虑了怎样合成语音库中没有拼接单元的情况。该方法能更好地利用自然语流的原始信息,提升了系统合成语音效果的自然度。This paper uses a Uighur speech synthesis system technology method using phoneme sequences with variable length as concatenation units, introduces the language character of Uighur, synergetic sound phenomenon which we must consider in the process of speech synthesis, gives the designing thought of multilevel speech database structure including sentences, phrases, words, syllables, phones and so on. We can find the concatenation units from speech database directly, in addition, we consider how to synthesis the texts that are not included in that speech database as well. This method can take advantages better of the original information of the natural speech and improve the naturalness of the synthesis speech.

关 键 词：语音合成 可变长音素序列 拼接单元 语料库 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]