And then go for the same song on different compilation albums and
compare.
Then try that on a different system/drive.
On this data material you test your ideas for normalization
(a method that should make all data comparable, e.g. by
transforming to a common intensity range like [0, 1],
a common sampling rate etc.) and distance.
Regards,
Marc
P.S.
Was it "Hunt on Red October" or maybe "The Abyss" where we had that
soldier who could listen to a sound and decide if it was a russian
submarine or a whale?
The U.S. Navy has those acoustical recognition systems for sure. :-)
Police folks too, to recognize voices.
P.P.S
Maybe even Music industry for the automatic scanning of MP3 sites..
Couple a web spider like "wget" with such an id system and you can
produce a nice list of copyrighted material despite whatever the
MP3 TOC says.
I am not sure if all MP3 friends would love for something like
that? :-)