摘要 |
PURPOSE:To generate a smooth synthetic voice less in deterioration by finding the frequency spectrum of a residual waveform read out of a storage means, converting the frequency spectrum of the residual waveform interpolated at a phoneme border by frequency components into a time waveform, and generat ing the voice. CONSTITUTION:A residue storage part 104 extracts the residual waveform used as a sound source for each phoneme according to input phoneme information, converts it into the frequency spectrum by Fourier transform, and sends it to a residue interpolation part 106. The residue interpolation part 106 generates the frequency spectrum of a sound source used of a phoneme or the connection part of a phoneme piece according to intepolation information inputted from an interpolation information input teminal 102 and a reverse Fourier transform part 107 converts it into the waveform a time area and sends it to a synthesizing filter part 110. The synthesizing filter part 110 composes a timbre waveform by using rhythm information, parameters, and the residue and outputs it from a voice output terminal 111. Consequently, the synthetic voice which has small deterioration even at the border part of the phoneme is generated. |