The Research Of Mongolian Speech Synthesis System Based On Verb's Affix And Stem

With the rapid development of computer technology and information technology, how to make more natural and friendly communication through man-machine interface becomes a research hot topic, in which the study of speech interactive way is one of the focal points of common concern. Speech synthesis is a very important technology in speech interaction, and it relates to many research areas of computer technology, linguistics, phonetics, speech signal processing, and psychology. Mongolian is one of the official languages in Inner Mongolia Autonomous Region, and also it is a language with a certain influence in the world. Thus the research and implementation of Mongolian speech synthesis technology has important significance and practical value for promoting the development of Mongolian information processing.This article focuses on the study of establishing the Mongolian TTS voice database based on the Mongolian language and voice characteristics. It also researches the content of the rhythm features extraction for Mongolian speech, rhythm modeling, rhythm adjusting, and speech synthesis. Major works are as follows:First, in the design of Mongolian voice database, through studying the characteristics of Mongolian voice, we chose verbs stem and affix,the additional composition of the norm and other whole phrases as basic unit for the speech synthesis. We collect speech basic unit from a big corpus too. Then we record and segment the speech basic unit,build up the voice database.Second, the extraction and modeling of Mongolian speech rhythm, through the analysis and statistics for the natural speech flow of Mongolian, we summarized the rhythm transformation rules of Mongolian language. Main expression lies in the changes of pause, duration, accent, and tone on the word, sentence level.Last, rhythm adjustment and speech synthesis, we achieve the rhythm adjustment of the synthesized speech using TD-PSOLA algorithm, which improved the naturalness of the synthesized speech of Mongolian language. When synthesizing speech, we used the algorithms of unifying with the soft concatenation and the hard concatenation, which aims to achieve the better synthesis effect.