| 摘要 |
PURPOSE:To control the lasting time of phonemes without lossing the feature of phoneme data by dividing the phoneme data into plural sections, allocating the number of frames corresponding to a lasting time length based upon the average pitch of each section and executing thinning or repeating in each section. CONSTITUTION:At the time of switching a phoneme in accordance with an input text, the original data of the phoneme concerned are read out. The read phoneme data are equally divided into four sections e.g. through objective pitches P0 to P3 or the like of the phoneme and the average pitch of each section is found out. The numbers of frames of respective sections for obtaining an objective time length TX are found out and respective frames are distributed in accordance with the original number of frames of each data for synthesizing a phoneme. The frames of each data are thinned or repeated in accordance with the rate of the distributed number of frames to the number of frames of each data such as CV, VC data, so that the lasting time length of each phoneme can be controlled without lossing its feature. |
