摘要 |
PROBLEM TO BE SOLVED: To obtain an encoded speech signal of high quality. SOLUTION: A speech signal which is inputted while divided at predetermined frame intervals is composed of a linear prediction parameter and a driving sound source signal corresponding to a linear prediction residue signal, and the speech signal inputted while divided at the frame intervals is further divided at subframe intervals, thereby representing the driving sound source signal, frame by frame, as encoding parameters including the pitch component of the speech signal, the pitch gain indicating its gain, a residue component other than the pitch component, and the residue gain indicating its gain. In this case, a variation pattern code book 16 wherein plural kinds of variation pattern of a pitch gain consisting of a specific number of dimensions are prepared is used to select the optimum variation pattern code vector from plural kinds of variation pattern code vectors for encoding, thereby interpolating the pitch gain. |