Scalable subband audio coding, decoding, and transcoding methods using vector quantization

A method for decoding a digital data stream representing an audio signal, in which at least one group of spectral components is calculated from the vector quantization indices included in the data stream, and the spectral components of the said group are combined during the construction of a decoded version of the audio signal, each component of the group being associated with a set of vector quantization indices used in the calculation of this component, in which the data stream has a variable rate, the method comprising a first phase in which the data stream includes sets of respective vector quantization indices for calculating a first group of spectral components and in which the correlations between spectral components of the first group are analysed, and a second phase, in which the digital data stream of lower frequency than in the first phase includes sets of vector quantization indices for calculating only a part of the spectral components of the first group, and in which at least one spectral component of the first group not belonging to the said part is calculated from the indices at least partially copied from a set of vector quantization indices read from the data stream and associated with a component belonging to the said part, for which a maximum correlation was determined in the first phase.

Method for coding an audio signal, in which at least one group of spectral components is obtained from the audio signal, and the digital output data stream is produced including vector quantization indices for at least some of the spectral components, wherein at least one pair of components exhibiting a maximum correlation out of the group of spectral components is selected, and wherein an identification code is included in the output data stream for each pair of components selected, and wherein at least some of the vector quantization indices are included in the digital output stream for only one of the components of each pair selected.

Method for transcoding of a digital input data stream representing an audio signal coded by successive time windows, to generate a digital output data stream of lower data rate, in which the formation of the output stream comprises, for a signal window represented by a number of ordered input bits, A, the copying of a number of input bits, A-B, and the suppression of a number of bits, B, from the input stream in order to reduce the data rate, and where the bits to be suppressed from the input stream are determined by information received from a decoder to which the output data stream will be routed.

Audio coder comprising means for producing at least one group of spectral components from an audio signal, means for calculating the vector quantization indices for at least some of the spectral components of the group, and a multiplexer to generate an output data stream including at least some of the calculated vector quantization indices, wherein the audio coder comprises, in addition, means of analysis for selecting at least one pair of components exhibiting a maximum correlation out of the group of spectral components, the multiplexer being controlled to include identification information for each pair of components selected in the output data stream, and so that at least some of the vector quantization indices be included in the output data stream for only one of the components of each pair selected.