| 摘要 |
The invention concerns a method wherein a decoder synthesises a set of successive frames of N samples of an audio signal from encoding data included in a digital flow received from the encoder and comprising, for only one subset of frames, data representing spectral amplitudes of the audio signal. The decoder determines for each of the frames of said subset, on the basis of the encoding data, the cepstral coefficients (cx_q[n]) representing some at least of said spectral amplitudes whereas for the frames not forming part of said subset, it interpolates said cepstral coefficients, and generates by means of said interpolated cepstral coefficients (cx[n-1/2] a spectral estimate of the audio signal which it transforms in the temporal domain to obtain the synthesised frame. |
