The Design And Implementation Of The Speech Synthesis System Of Minnan Dialect

With the popularity of intelligent devices,people put forward higher demands on speech synthesis system.Dialects make people feel cordial and friendly,and help to narrow the relationship between mankind and machine.Meanwhile,our country also strongly advocates the protection of dialect as it is one of intangible cultural heritage.Therefore,it is of great significance to study the speech synthesis system of Min-nan dialect.At present,a mature speech synthesis system is the speech synthesis system based on hidden Markov model.This paper will design and implement Min-nan dialect speech synthesis system considering the actual situation of Min-nan dialect,and make appropriate improvement.The main work is as follows:(1)Phonetic study of Min-nan dialect pronunciation,and this will lay a good foundation for speech synthesis system;(2)There are many polyphonic words in Min-nan dialect.If cannot correctly distinguished,they will seriously affect the accuracy of speech synthesis system.This paper will use association rule to recognize polyphonic pronunciations;(3)The speech synthesis system based on hidden Markov model and its improved average tone model are studied and applied directly to the speech synthesis of Min-nan dialect.The naturalness of the output speech is not high.This paper will analyze the actual situation of speech synthesis of Min-nan dialect,and use the speaker adaptive training method to improve the average tone model proposed.(4)Based on the work above,the prototype system of Min-nan dialect speech synthesis will be realized,and the component modules and prototype system are preliminarily tested to verify the practical value of the system.