Le 28/05/2013 06:12, Ivica Ico Bukvic a écrit : Hello LAU list, Hello Pierre, I was surpised and yet seduced by the reading of the page (see above) relating to an audio coding format from Pierre Jocelyn. It is written in french. But found the idea imaginative enough to be worth shortly described in this list too. So, as far as i understood, the idea is to split the signal into a real-time (un-derminable) part and a predetermined part that can be extracted and shared separately to lossily reproduce the original audio signal. The implementation was to encode the signal in a stream of tuples of 3 integer values where the third one is a key of a non-resonant filter parameter, the second value is a step response duration, and the first value is the signed step amplitude. The stream is very light then, but the dictionnary of filters can be heavy. What is interseting is the possibility to share families of dictionnaries for a certain kind of sounds (speech, music, ...) appart from the stream, so that hardware could have presets dicts and only the stream of variable-length steps is transmitted. I find that fun, but i wonder if GSM phones does already use that kind of principles with their adaptative sound repdroduction algorithms. Correct me if i'm wrong... Cheers, Benoît.