摘要 |
PURPOSE:To find a more accurate impulse response waveform even for a voice which has a short pitch cycle like a female voice. CONSTITUTION:A voiced sound is generated by convolutional integration between an impulse train expressing a vocal chord part and an impulse response waveform representing a glottis part. The impulse response waveform which becomes data for synthesis is found by this voice synthesis system. At this time, the frequency characteristic for the glottis part is calculated from an analytic parameter found by a voice characteristic generation model which inputs the impulse train to find the impulse response waveform of the glottis part. |