Proposal for new music analysis algorithm with very practical uses inside Amarok

[Manuel Amador (Rudd-O)]

Replies interspersed:

El Lunes 13 Octubre 2008, Sven Krohlas escribió:
>
> I guess most of us don't know much about music similarity recognition
> (including me), so there aren't many replies here, yet. For me there is
> one important question: is it possible to use one algorithm (and so one
> fingerprint) for your use cases and Sorens?
> Calculating a fingerprint takes a relatively long time, I guess, and
> collections grow really fast these days.

Fingerprinting my entire collection (17000 songs) according to my calculations 
requires about one day and possibly a a few more hours.  That can be done in 
idle time (nice 20) so it's not a problem.  Correlating all songs can take 
another two days.  This process can take as long as needed and can be 
augmented by the use of anonymous distributed computing without user 
intervention (I'm devising precisely that at the moment).

Soren's algorithm does not appear to work like mine at all.  His uses 
frequency correlations, mine uses time-series correlations among corresponding 
frequency bands.  As such, they deliver entirely different results.

We don't have to choose one or the other.  We can vectorize both algorithms 
just fine.
-- 

	Manuel Amador (Rudd-O) <rudd-o at rudd-o.com>
	Rudd-O.com - http://rudd-o.com/
	GPG key ID 0xC8D28B92 at http://wwwkeys.pgp.net/

Now playing, courtesy of Amarok: 19 - Hanson _ Where s The Love
If A = B and B = C, then A = C, except where void or prohibited by law.
		-- Roy Santoro

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part.
Url : http://mail.kde.org/pipermail/amarok-devel/attachments/20081013/8345639a/attachment.sig