摘要 |
PURPOSE:To solve the problem that an input audio signal has the waveform distortion increased in the band by the outband component of a driving signal at the time of encoding the telephone voice, where the input voice has the band limited, to a linear forecast code. CONSTITUTION:The input telephone voice is set to zero phase by a phase equalizing filter 4, and the same band limiting as the input telephone voice is performed by a band pass filter 5 to obtain a signal sb(t). The signal set to zero phase is subjected to linear forecast analysis, and a forecasting filter 10 is controlled by a forecast coefficient ai. A quasiperiod pulse train where fluctuation of the pitch of the input voice is limited is generated by a pulse train generating part 7, and this pulse train P(t) has the band limited by the same band pass filter 8 as the band pass filter 5, and the output is supplied as the driving signal to the forecasting filter 10 through an amplitude control part 9, and voice synthesis is performed. Each pulse amplitude is calculated in a pulse amplitude calculating part 13 so that the square error between a synthesized signal s'(t) and the signal sb(t) is minimum, and the amplitude control part 9 is controlled. |