| 摘要 |
PURPOSE:To obtain the rule synthesized voice which is less deteriorated in sound quality by executing time length control at the time of smoothing the fluctuation of pitch extraction and executing pitch control. CONSTITUTION:The pitch of the unit voice to be processed is inputted from a unit voice input section 15 is sent to a pitch smoothing section 2 in a signal line 1. The pitch smoothed by the pitch smoothing section 2 is sent to a pitch controlled variable calculating section 3. The residual signal is subjected to the pitch control by zero cutting the discretization from the pitch controlled variable and the pitch of the input voice inputted by a signal line 7 in the pitch control section 5. On the other hand, the spectral parameter of the unit voice to be processed, i.e. alpha parameter in this case are sent by a signal line 9 to a unit voice connecting section 12. The residual signal and the alpha parameter are controlled in time length in a time length control section 13 and are sent to a voice synthesizing section 8. The voice synthesis is executed in the voice synthesizing section 8 and the synthesized voice is outputted from a voice output terminal 10. |
