I have a collection of CDs converted to MP3s when I didn't have full time net
access, and thus are missing ID3 tags from CDDB. I also have a some
miscellaneous MP3s downloaded from the internet. Since CDDB and friends work
from the CD TOC and track lengths, it's not possible to automatically identify
an MP3, and it's not possible to, say, identify an MP3 by the first few bytes
since different encoders could produce different outputs.

Has anyone heard of any way to identify a music track automatically? The way
I would envisage it would have to work would be along the lines of a speech
recognition program that converts the first few seconds to its raw format
and perhaps performs a FFT to convert it some fundemental characteristics
and looks them up via a fuzzy matching mechanism. Hardly an afternoons, work...

At the moment, I'm feeding my CDs back in and running "cdcd tracks" (on linux)
and scripting "mp3id3" (from mp3tools) to add in the ID3 info, followed by
moving the mp3 to /music/genre/artist/album/songname.mp3. Then I'm running
the linux "emptool" program via an expect script to lable the tracks on the
empeg.

There's got to be an easier way!

Cheers,

Paul