To enhance the performance of text-to-speech systems (TTS), we generated various speaking styles. As the first trial, we selected three speaking styles appropriate for task specific texts, and analyzed their characteristics. Large differences were found in the 1st and 3rd formant frequencies, fundamental frequency range and phrase height assignments, speaking rate, and segmental duration of phrases followed by pauses. Rules for making the characteristics of the speaking styles were integrated into a conventional TTS system, and listening tests confirmed the good performance of the rules.
|Number of pages||7|
|Journal||NTT R and D|
|Publication status||Published - Dec 1 1996|
ASJC Scopus subject areas
- Electrical and Electronic Engineering