[PD] audio fingerprints

[David Sabine]

Hello all,

I'm wondering if anybody has explored the process of creating and retrieving
audio "fingerprints".  My experience with such a process is very limited, so
please forgive me if my jargon is un-informed.

What I'm interested in is simple:

1.  record a sound.
2.  generate a "fingerprint" - perhaps based on peak frequencies, wave
shapes, etc.?
3.  test that fingerprint against a database/collection of other
fingerprints
4.  return a value dependent on whether or not a match was found and perhaps
a value (percentage?) representing how reliable the 'match' would be.

I've taken a look at [choice] in the "extra" folder but it doesn't seem
immediately applicable to this project. ?  Perhaps I'm misunderstanding this
object and its usage.?

The eventual goal would be to record a performance or excerpt of a
performance and be able to determine which song is being played.

Many thanks,
Dave S