Process for reducing data for transmission or storage of digital audio signals from several interdependent channels, in which blocks of samples of signals in the respective channels are transformed from time domain into a frequency domain representation, whereby a value is determined for each of a plurality of frequency components in each channel, and the values determined for the respective frequency components are coded, taking account a masking threshold determined by means of a psychoacoustic model, said process comprising the steps of:
- determining a first data rate necessary for separate coding of signals in each of the respective channels, said first data rate being determined collectively for all frequency components of signals in said respective channels;
- determining a second data rate necessary for joint coding of said signals in the respective channels, said second data rate being determined collectively for all frequency components of signals in said respective channels;
- comparing said first and second data rates;
- performing joint coding of said signals for all frequency components of the respective channels so long as the data rate necessary for joint coding of said signals does not exceed the data rate necessary for separate coding by a predetermined threshold value;
- and
- performing separate coding of said signals for all frequency components of the respective channels, when the data rate necessary for joint coding of said signals exceeds the data rate necessary for separate coding by at least said predetermined value.