Resynthesis the correct tone utterance of a sentence from a corresponding arbitrary tone utterance

The current paper presents a system that generates tones automatically from utterances and sentences. Users were asked to speak a sentence displayed on the screen with any arbitrary tone sequence (level and falling tones is preferred). The system records the utterance, determines the correct tone sequence from the sentence, determiners syllable boundaries by force-aligning the utterance, generates a new Northern Vietnamese tone F0 contour by Fujisaki model, then resynthesize the utterance with the new tone sequence. The system could be applied in teaching foreigners to speak Vietnamese with the correct tones or converting utterances of a Middle Vietnam or Southern Vietnam speaker to Northern tone system.