[music-dsp] lossless pcm compression

robert bristow-johnson rbj at audioimagination.com
Mon Feb 28 11:08:21 EST 2005


on 02/28/2005 04:30, Knut-Inge.Hvidsten at ffi.no at Knut-Inge.Hvidsten at ffi.no
wrote:

> Anyone into the lossless codecs?

i used to have some interest in it.

> It seems that they offer about 2:1
> compression. I thought that was a pretty low ratio? Is that en expression for
> the entropy of music, or limitations in the algorithms?
> 
> It would seem to me that using a predictor of some order while exploting
> symmetry in a) positive/negative half-waves b)left/right signal a factor near
> to 4:1 plus predictor reduction would be possible??

i think that applying LPC (delta modulation is the simplest LPC) to the
audio which results it a bunch of samples representing the difference of the
LPC guess and the actual sample, and then applying Huffman coding (with a
predefined code book that does not have to also be transmitted, but some
parameters of the code book as well as the LPC filter *do* get transmitted)
to that difference is about as good as it gets for lossless.  the reason why
is that the result will be white and huffman will get rid of any redundancy
of the signal (or difference signal) hanging around zero more often than
anywhere else.

-- 

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."





More information about the music-dsp mailing list