規則合成における様々な声質の実現の試みについて  [in Japanese] Tentatively about the achievement of various characters voice quality in the rule synthesis.  [in Japanese]

筆者らは声道模擬型の規則音声合成器の研究をしている. 今回人間の放射音声/a/から求めた逆フィルタ波形を1波形に切り出さないで連続したまま音声合成器へ入力した. すると, その人の個人性を保存した肉声レベルの/aiueo/を合成できた. このため, この方法を使えば, 数式でのモデル化が困難な声質も実現できる可能性がある. また, 筆者らの合成方法はターミナルアナログ型であるため, 人間の発声機構の理解に役立つと思われる. 今後は様々な性質を実現する.

Authors are researching the rule voice synthesis machine of a vocal tract simulation type. This time, we use a reverse-filter crimps which calculated from radiation voice/a/as driving sound source of the voice synthesis machine was. Then, it was able to synthesize/aiueo/at the voice level by which the person's individual was preserved. Therefore, there is a possibility that various voice quality that modeling by the expression is difficult can be achieved if this method is used. Our aim is achieve various voice quality with the terminal rule voice synthesis machine in the future.