| 摘要 |
<p>PURPOSE:To improve the quality of a reproduced voice by reducing the error rate of the voice condition discrimination of a voiced/voiceless discriminator in a variable bit rate.low bit rate voice code decoder. CONSTITUTION:Voice signals b1 which are framed by a framing device 11 are wavelet-converted by a wavelet converter 12. A result c1 is computed by a wavelet conversion result energy computer 13 and the energy value for every frame is inputted to a condition discriminator 14. The energy threshold values for sound/silence segment discrimination which is beforehand set by the discriminator 14, periodicity discrimination threshold values and threshold values for checking the presence and absence of rapidly formant changes are compared, discrimination is made for any one of a voiced section, a voiceless section, a voice presence/absence transition section and a silence section and a corresponding flag e1 is outputted.</p> |
