Research On Key Technologies Of Tibetan Text-to-speech Conversion System

Nowadays with the development of information technology and approaching of the main data era,the demand for applying computers into the information process is increasing.Dealing with language and text by means of computers has been a hot theme in the field of computer applications.The birth of TTS is an important achievement of handling the language and text by computers.It handles text information and converts it into a fluent and understandable voice signal,outputting it accordingly,which plays an important role in daily communication,human-machine interaction,artificial intelligence and other fields.The research on TTS began in the 1990 s and successfully developed a translation system containing a relatively wide range of languages,such as Chinese,English,French,etc.In China,the development of TTS for some minorities' languages also made remarkable achievements.Tibetan is an important part of the national system in our country and mainly distributed in Tibet,Qinghai,Sichuan,Gansu and other provinces.With the development of science and technology and the construction of national road engineering,more and more people are intended to visit these places,thus involving the Tibetan culture with other ethnic one.In the hope of making Tibetan and other ethnic groups more efficient to communicate in cultural exchanges,the study of the Tibetan language and language conversion system is of great significance and wide application prospects.The main research work in this paper is as follows:1.On the basis of the study of Tibetan structure and writing grammar,Tibetan syllables as basic elements are chosen by the author.By selecting and extracting the features of each Tibetan character in Tibetan syllables,the Tibetan syllable feature vectors are constructed.Based on Tibetan grammar,the location of every syllable in Tibetan language is discussed and a special base syllable bank is also established,which enables the system to accurately predict the location of base words in Tibetan syllables.2.After the analysis of the application,the Tibetan syllable feature vector library based on Tibetan syllable structure and the Tibetan syllable feature vector library based on the number of Tibetan syllable characters are constructed respectively.3.The problem of text recognition is solved by searching syllable feature vectors in syllable bank.4.According to the actual application of speech synthesis,two Tibetan phonetic banks are set up,which are Tibetan syllabic speech fragments.Moreover speech banks based on mixed excitation and LPC speech coding are compressed.Based on the above research results,a Tibetan text to speech system is constructed.Through experimental verification in MATLAB environment,the system can work normally when the feature vector database is sufficient.