| 摘要 |
PURPOSE:To eliminate the unmatching of syllable data when it is coupled by analysizing a waveform, generating syllable information such as CV information, VC information, and common V part information, polymerizing each information by performing weight mutually between transition intervals, performing waveform analysis, then, finding each parameter. CONSTITUTION:The ad interval of the latter part is stored from the V part(vowel example, 'a'(vowel in Japanese language)) of an information wave, and a bc interval is set as a common interval, and the ad interval and a cd interval as the transition intervals. In the waveform interval eh of a CV waveform KA, an fg interval is set as the transition interval, and is formed with the same width as that of an interval ab. Such processing that the waveform of the common and transition interval ab of a waveform aA is weighed by 0 at a point (a) and by 1 at a point (b) are performed, and the weight of 1 at the point (f) and the weight of 0 at the point (g) of the interval fg of the CV waveform are performed, and both waveforms are superimposed at the intervals ab and fg, the a CV information waveform KA is generated. Similarly, weighing coupling by a transition period is applied to the VC information. By generating the syllable information by analyzing the CV waveform, a VC waveform, and a common V waveform obtained in such way and extracting respective parameter, no matching error occurs, and no abnormal tone occurs among the vowel, a nasalization vowel, and a consonant, which enables a smooth synthetic sound to be generated. |
