Break Predicting Methods Using Phonetic Symbols Combined with Accents Information in a Japanese Speech Synthesizer

일본어 합성기에서 악센트 정보가 결합된 발음기호를 이용한 Break 예측 방법

  • 나덕수 ((주)보이스웨어) ;
  • 이종석 ((주)보이스웨어) ;
  • 김종국 (숭실대학교 정보통신 전자공학부 소리공학연구소) ;
  • 배명진 (숭실대학교 정보통신 전자공학부 소리공학연구소)
  • Published : 2007.06.30


Japanese is a language having intonations, which are indicated by the relative differences in pitch heights and the accentual phrases (APs) are placed according to the changes of the accents while a break occurs on a boundary of the APs. Although a break can be predicted by using J-ToBI, which is a rule-based or statistical approach, it is very difficult to predict a break exactly due to the flexibility. Therefore, in this paper, a method which can enhance the quality of synthesized speech by reducing the errors in predicting break indices (BI), are proposed. The method is to use a new definition for the phonetic symbols, which combine the phonetic values of Japanese words with the accents information. Since a stream of defined phonetic symbols includes the information on the changes in intonations, the BI can be easily predicted by dividing the intonation phrase (IP) into several APs. As a result of an experiment, the accuracy of break generations was 98 % and the proposed method contributed itself to enhance the naturalness of synthesized speeches.