摘要 |
PURPOSE: To output natural and varied synthesized voices rich in diversity by preliminarily storing plural tables with respect to accent and multiplying a multiplication coefficient based on a single table with respect to intonation. CONSTITUTION: An accent command generation circuit 3 generates an accent command synchronized with the change of a tone feature vector, which is determined by a tone feature vector selection circuit 2, based on an accent symbol in an input text. An accent component operation circuit 4 selects an accent component from plural accent component tables based on the step height of the accent command and calculates a pertinent accent component value based on the current frame number and the difference between start and end frame numbers. A phrase command generation circuit 5 generates a phrase command, which is synchronized with the change of the intonation vector, based on a delimiter symbol in the input text. That is, a phrase component table for a reference impulse height is preliminarily stored, and the value of a ratio of the input impulse height to the reference impulse height is taken as the multiplication coefficient, and the phrase component table value is multiplied by this coefficient. Thus, various semantic contents are easily expressed. |