| 摘要 |
A predicted residual signal is calculated from a current input speech signal and a past input speech signal, and a cross-correlation between the predicted residual signal and the past input speech signal having one speech sub-frame length stored in a first code book is calculated. In cases where the current input speech signal has no local peak, the cross-correlation becomes high, so that a synthesized speech signal is generated from the past input speech signal stored in the first code book or a predetermined sound source signal having one speech sub-frame length stored in the second code book. In contrast, in cases where the current input speech signal has a local peak, the cross-correlation becomes low, so that it is judged that a function of the first code book is depressed. In this case, a synthesized speech signal is generated from a group of short-length sound source signals having a total length equal to one speech sub-frame length stored in a short-length signal code book. Therefore, even though the current input speech signal suddenly has a local peak, because the synthesized speech signal is generated from the short-length sound source signals respectively having a speech length lower than one speech sub-frame length, the local peak can be expressed by the short-length sound source signals, an appropriate exciting sound source signal similar to the current input speech signal can be determined, and the synthesized speech signal can be adequately obtained.
|
