| 摘要 |
PROBLEM TO BE SOLVED: To improve the quality of synthesized voice by a rule by making several subjects listening a synthesized voice made by using several sets selected at random out of syllable balance meaningless word sets, measuring clearness and understandability of rule synthesized voice generated by a rule synthesizing device, and verifying the cause of deterioration of a voice. SOLUTION: The occurrence frequency of all two syllable chains and three syllable chains of a population 1 is calculated. Words in a meaningless word set is constituted so that occurrence frequency of two syllable chains and three syllable chains in the meaningless word set is approached to this relative frequency. Words are generated by deciding possibility of connection of two syllable chains by kinds of three syllable chains of the population 1 and deciding the number of times of possibility of connection by the ocurrence frequency after this processing. A syllable balance meaningless word set made in the above way is divided into several sets having numbers with which listeners are not tired, and several sets are selected out of them at random. This is used for evaluating a synthesized voice. |
