摘要 |
A voice encoding method and apparatus of high quality, which has little local noise generation of decoded sounds. The voice encoding apparatus comprises: drive vector generating means (13) for generating a plurality of drive vectors; a first strain calculating unit (23) for calculating a waveform-related strain as defined for each drive vector between a signal to be encoded and determined from an input voice and a synthetic vector determined from the drive vector, as first strain; second strain calculating unit (24) for calculating a second strain, as different from the first strain and as defined for each drive signal between the signal to be encoded and the synthetic vector determined from the drive vector; an evaluation value calculating unit (29) for calculating a predetermined searching evaluation value by using the first strain and the second strain; and searching means (20) for selecting the drive vector to minimize the searching evaluation value, thereby to output the code which is made in advance to correspond to the drive vector selected.
|