| 摘要 |
PROBLEM TO BE SOLVED: To provide a voice synthesizing method and its device capable of synthesizing voice higher in naturalness than in the past without requiring much data to be statistically processed. SOLUTION: Vector series composed of rhythm vectors each made up of an interval period and power (amplitude), are extracted from a voice and previously stored in a rhythm-vector-series data base 10. The voice is made by continuously phonating typical syllables of voiceless fricative sound and voiced plosive sound, classified by articulating positions and articulating methods. Rhythm setting 15 is performed by the selection of a rhythm vector series at a rhythm selecting part 11. The selection of the rhythm vector series at the rhythm selection part 11 is performed so that syllables and phonemic chains of the voice are equal to those of a voice to be synthesized in mora number and accent pattern and in phonemic class. Meanwhile, segment selection is performed by a segment selecting part 16. With respect to a character string for which reading of each rhythm-set and segment-selected character is converted into a voice, a segment deforming part 17 performs segment deformation by changing the basic frequency and amplitude of segments according to their accent and intonation. |
