Modeling the timing characteristics of different speaking styles

To design models for synthesising speech with natural rhythm, we investigated the relationship between the speed of speaking and the duration of phonemes and pauses. The purpose of this research is to predict duration from texts on the basis of syntactic analyses. For data, we asked three subjects to read a well-known fairy tale at different speaking rates. Using this data, we compared the durations of the different speaking styles and show how speech timing structure changes according to the styles of speaking, based on syntactic analyses.