Abstract: Using a group of spectral components from an audio signal, a coder produces a digital data stream (&PHgr;) including the quantification indices (QE0m) of at least some of the spectral components (E0m). The coder, or a transcoder situated downstream, selects at least one pair of components exhibiting a maximum correlation out of the group of spectral components, and includes in its digital output data stream identification information for each pair selected. At least some of the quantification indices are included in the output data stream for only one of the components for each pair selected. The decoder will use these to obtain the suppressed indices for the other component of the pair.

Abstract: On the basis of a portion of the signal, a coder selects a parameter vector belonging to a reference library containing 2Q vectors each designated by an address of Q bits. This coder, or a transcoder located downstream, forms a digital data stream (&PHgr;) containing an index deduced from the address of the vector selected. For each rate value corresponding to an integer p≧0, the index contained in the digital data stream is formed of Q−p bits which, completed by p bits of predetermined positions, define indices of Q bits representing a group of 2p addresses including that of the vector selected from the reference library. The decoder receiving the digital data stream, or a transcoder located upstream, is capable of reconstructing appropriate parameter vectors on the basis of the truncated index. Thus very fine steps are obtained in the adjustment of the transmission rate.