| 摘要 |
PURPOSE:To obtain a proper sound source even if voice/no-voice information can not be corrected by enabling a synthesis part to calculate voice/no-voice information from voice spectrum information by using the same algorithm with an analysis part, compare it with received information, and correct the information. CONSTITUTION:The voice spectrum information C(14) is sent from a separator 3 to a voice/no-voice decision device 9 to calculate reduced power PL. The power PL is compared with a voice/no-voice threshold value PLTH to calculate the voice/no-voice IXP. Then the information IXP from the decision device 9, the information IVN from the separator 3, and pitch information IPN are sent to a corrector 10, which substitutes held last-frame voice/no-voice information IVB with the information INV as correcting operation only when IPN not equal to 0 and 'IVN not equal to IXP'. Then a sound source switch 6 is switched to a noise generator 5 when INV = 0 or to a pulse generator 4 when INV not equal to 0 to obtain the proper sound source for a synthesizing filter device 7. |
