Audio Fingerprint

Audio fingerprints are identifiers for audio files based on the contained audio data.

Up to early 2006 MusicBrainz has used Relatable's TRM, which is an algorithm matching files by audio fingerprints. However, we have been GettingRidOfTRM, because we have reached the limits of scalability, and performance is not as good as we would like. Because the server is closed-source, we have no way to fix either of these problems.

For a short while we debated whether we should attempt to build our own open source FutureProofFingerPrinting technology, but nobody was there who could have done the work.

Since March 2006 MusicBrainz collaborates with MusicIP who operate the MusicDNS service. Their (proprietary but much better) PUIDs have been integrated into the Picard 0.7.0. See PUID and HowPUIDsWork for details.

Known Audio Fingerprinting Systems

Open source systems:

• The fingerprint in Kurt Rosenfeld's FDMF.
• Freetantrum songprint
  There is a project called freetantrum on Sourceforge. It seems to be a dead project (its home page was replaced with an advert for unrelated things in 2001), but it may be worth investigating and resurrecting the code they produced.
• MusicURI, a project trying to combine audio fingerprinting with text based search. More info on project homepage. The GPL source code is available at SourceForge.
• A possible FutureProofFingerPrint algorithm that we could code has been Proposed by Geoff Schmidt, formally of Tuneprint.
• libFooID, an audio fingerprinting library used by, and developed for foosic.

Commercial systems:

• Relatable's TRM
  The system MusicBrainz originally used. The use of TRM has since been removed from MusicBrainz.
• MusicIP's PUID (formerly MusicMagic/Predixis)
  The PUID technology has been integrated into the MusicBrainz since March 2006 and into Picard since version 0.7.0 released in July 2007. See their patent application for details on the technology.
• AudibleMagic
  This article compares business models for Philips & Audible Magic
• Philips (also licensed by SNOCAP)
  http://ismir2002.ismir.net/proceedings/02-FP04-2.pdf http://ict.ewi.tudelft.nl/pub/doets/paper027.pdf This presentation (in PDF) has a few slides on the Philips system as well:
• USC
• AudioID:
  On the Fraunhofer IDMT site, site of the AudioID company
• Shazam have their own proprietary music fingerprinting system, currently offered directly to users over the telephone.
• And possibly most ridiculously of all: The Song Tapper