摘要 |
PURPOSE:To enable waveform synthesis characteristic to respective phonemes by storing phoneme niece waveforms of various sizes corresponding to the phonemes or the kinds of the phonemes, and deforming and connecting the waveforms. CONSTITUTION:An input part 10 for a phoneme series, etc., inputs a pitch frequency based upon the phoneme series, each phoneme time length, and the waveforms forming the phonemes. A kind waveform selection part 20 extracts the phoneme or the kind of the phoneme and selects a phoneme piece waveform among, the stored phoneme piece waveforms according to the inputted phoneme or phoneme time length. Then a kind-classified waveform deformation part 30 deforms the selected waveform by using a window having window length and a window shape corresponding to the kind of the phoneme or the time length. Then a kind-classified waveform connection part 40 connects the deformed waveform according to the kind of the phoneme, time length, or a pitch frequency. A natural speech with articulation can be synthesized by this waveform editing process. |