Expressive Text-to-speech System On Mandarin

Expressive text-to-speech system has a wide variety of applications. Compared withgeneral speech synthesis for Chinese, this paper focuses on prosody and intonation.In general, prosody is described from three aspects, accent, pause and speaking speed.Accent is the words which sounded more prominent than others in a sentence. It can bestressed by raising the fundamental frequency appropriately and increasing the amplitudeof the word. Pause refers to the intermission between sentences or words, it can beachieved by interpolating some frames which parameter value is zero in the correspondingposition. Speaking speed refers to the length of each syllable and the degree of couplingbetween syllables. It is controlled by copying or deleting some frames in specifiedlocation.Since mandarin is a tonal language, intonation is significant in the synthesis.Intonation affects throughout the sentence, it is reflected on the pitch contour of the wholesentence, especially at the end syllables of the sentence, it ignores the tone of each word.There are four intonation patterns in mandarin according to the different tone and emotion,they are rising intonation, falling intonation, flat intonation and sinuate intonation. Extractthe fundamental frequency of the prepared speech corpus by STRAIGHT. And normalizeall fundamental frequency. Use polynomial fitting function to model the fundamentalfrequency track for each intonation pattern. There are three ways to model the intonation:Mean Model, Single Gaussian Model and Gaussian Mixture Model. And then apply theintonation model to convert one pattern to another.Use STRAIGHT to synthesize at last. And it can be seen from the experimental results,the proposed method can achieve a good quality on the conversion of tune. What’s more,the expressive text to speech system can highly improve the naturalness of the speech.As a whole, prosody and intonation play important roles in mandarin speech. They areessential in the speech synthesis system.