| 摘要 |
<p>The invention concerns a method wherein the encoder estimates a fundamental frequency (F0) of an audio signal, and determines a compressed higher envelope of the spectrum of the signal by interpolation of the spectral amplitudes associated with multiple frequencies of the fundamental frequency with application of a spectral compression function. It transforms the compressed higher envelope in the cepstral domain to obtain cepstral coefficients (cx_sup), and includes in a digital flow transmitted to the decoder quantization data of said cepstral coefficients. The cepstral coefficients are transformed by liftering in the cepstral domain before they are quantized. A value of the spectrum module of the audio signal is recalculated at at least multiple frequency of the fundamental frequency on the basis of the transformed cepstral coefficients (cx-sup q), and the liftering is adapted so as to minimise the module difference between the audio signal spectrum and at least a recalculated module value.</p> |
