Closed oleander closed 9 years ago
Echoprint's audio fingerprints are composed of lots of small hashes ("codes") representing only a few hundred milliseconds of audio. That SQL query you pasted is counting the number of matching codes grouped by track, where the track with the most code matches would be considered the closest match.
Aha, I get it. Thanks for your respons.
I took a look at the implementation and only saw full fingerprint matching against the database, no partial matching. That that would mean that the echoprint client have to generate an exact replica of the fingerprints already existing in the database, or I'm I wrong?