| 摘要 |
PURPOSE:To improve tone quality more effectively when a bit race is lowered by classifying sound source signals by predetermined kinds and using a pitch prediction multipulse train and a code book or increasing the number of multipulses. CONSTITUTION:Calculating circuits 515 and 520 find a pitch parameter and a spectrum parameter from input discrete speech signals and a decision circuit 575 classifies them by the predetermined kinds; when characteristics are improved by the use of the multipulse train or code book which is found by pitch prediction, the sound source signals of the speech signals are sent while represented by using both the code book and multipulse train. The sound source signals are decoded by using the code book and multipulse train according to the kinds and a composite speech signal is outputted through a composing filter circuit 760 by using the spectrum parameter. |
