摘要 |
PROBLEM TO BE SOLVED: To automatically generate a group of speech messages having uniform rhythmic features at low cost. SOLUTION: A rhythm normalization system is equipped with a rhythm time-series data input means 1 of entering time-series data of a rhythm before conversion which is extracted from speech data, a representative parameter calculating means 2 of calculating a representative parameter representative of a feature of the rhythm from the entered time-series data of the rhythm, a target parameter input means 3 of entering a target parameter determining a rhythm as a conversion target according to the kind of the inputted time-series data of the rhythm, a rhythm converting means 4 of calculating a variation rate of the rhythm from the representative parameter and target parameter and normalizing the rhythm time-series data according to the variation rate, and a conversion data output means 5 of outputting the rhythm time-series data having been normalized by the rhythm converting means. COPYRIGHT: (C)2005,JPO&NCIPI
|