>
> well, i *am* a mathematician, but i must admit that all the above
> mentioned things mean very little to me. however, it might not be too
> difficult for me to get into this stuff. can you suggest some reading
> to get started?
Sorry, *that* message was to scare off the easily impressed :-) I'll describe
more carefully from here on.
> > that makes use of spectral features with widths naturally based on the octave
> > they appear in, 4:1 lossless should not be too difficult to acheive.
>
> does that imply you think we can get 4:1 lossless on average audio
> sources?
Closer to it, anyway. Of course, this *still* depends on the amount of real
entropy in the signal, but most of the perceived entropy in a signal encoded
using an LPC coefficient+residual scheme is simply due to a poor fit (a
consequence of the nature of the filter).
> that would be exciting, if not sensational! i have my doubts,
> however... what about the cymbals you mentioned in an earlier posting
> for example? basically, they produce pure random noise, don't they?
Actually, cymbals produce strong and weak peaks in a pattern that mostly appear
random. They're still a worst case, but LPC could be doing better than it is
doing :-)
What remains to be seen is if what I want to do is even possible. I'll also
answer the previous mail that asks specific questions I'm sure you'd also like
answered...
Monty
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
