From: V.Lanard <vl...@ya...> - 2000-10-27 00:47:47
|
-------------------------------- 1. Which part of a song is the Songprint ID really based on? -------------------------------- I've been experimenting with sp_example to determine how small of an MP3 will successfully match an entry in the Songprint database. What I've found is that the first 30 seconds (or 400K) of a song seems to work pretty consistently, but any other portion of the same song is not recognized, regardless of length. I don't know whether this is something specific to sp_example or the underlying algorithm and I haven't started looking at the code. This leads me to wonder whether 30 valid seconds of a song followed by a cuckoo's egg would fool the database? Also, for any clip of a song to be recognized, that exact clip would have had to be entered into the Songprint database by someone, right? So seconds 10-40 of a song would have a different key than seconds 0-30. And judging by my testing of sp_example, the entire song's key matches the key from seconds 0-30 but not 10-40 or 0-10 or any other clip. This seems like it will lead to a situation where any imperfect encoding of a song (ie. a song missing the first few seconds) would cause the Songprint database not to recognize the song unless someone had submitted a key from exactly that imperfect version. -------------------------------- 2. Is there any easy way to capture the Songprint key locally from sp_example? -------------------------------- I'm trying to identify the Songprint ID generated by sp_example so that I can use it to query Songprint separately from the key extraction process. Is there an easy modification that I can do to sp_example to allow me local access to the key? Thanks! V.Lanard |