| 摘要 |
PROBLEM TO BE SOLVED: To perform a speaking speed conversion by a PICOLA (Pointer Interval Controlled OverLap and Ad) system so as not to generate a frame, from which speech waveform data is not outputted, in a streaming processing. SOLUTION: The speaking speed conversion method comprises: a step S2 of holding the input speech waveform data for one frame component into an input buffer 5; a step S6 of acquiring the input speech waveform data of a data length necessary for determining a basic cycle of the speech waveform and holding the same as data for searching the basic cycle; a step S11 of determining the basic cycle of the speech waveform from the data for searching the basic cycle and freshly generating the speech waveform data extended or compressed with the speech waveform, based on the basic cycle and the speaking speed conversion magnification; and a fifth step S14 of outputting only the number of samples, obtained by dividing the number of samples of the one frame in the speech waveform data generated by the fourth step S 11 by the speaking speed conversion magnification and holding the remaining speech waveform data in an output buffer 8. COPYRIGHT: (C)2006,JPO&NCIPI |
