My brother swears this is true. Lately he has been transfering some of his old vinyl albums to his computer using his nifty new USB equipped turntable. He divides the tracks into separate files then puts them into iTunes. Now here’s the weird part. He claims that iTunes checks the files and provides the artist and title information! Now I can understand that it does this with digital music which probably carries some signature in it’s code to clue the iTunes computer as to what it is, but how does it recognize music taken from vinyl? Does it have the ability to compare sound waves it receives with sound waves it has in its data base?
I have a hard time believing this.
This line is equivalent to when a fisherman starts out a story with, “No shit, this really happened”.
Hmm… maybe I should take some of the music I’ve recorded with my MP3 boombox’s line-in feature and try importing it into Itunes. Some of them are songs taped off the radio that I haven’t been able to identify ~10 years later.
I know one of the cell phone companies has been lately advertising a feature that allows you to key the phone to listen to an unfamiliar song, identify it, and let you download it into the phone. So the tech exists.
Here’s a description of the service I found with Google. Doesn’t mention iTunes, but I can easily believe they’ve got the same ability.
It’s actually not that complicated. Gracenote (the company that provides Itunes’ media recognition, at least for now) announced some time ago that they were trying to support recognition of files ripped from vinyl. I’m assuming it’s digital file recognition based on the mp3 (or format of choice) that results from ripping…which is exactly how they do recognition on digitally sourced single tracks from CDs anyway. The technology is well established.
As far as I know, TOC (table of contents) data is used only for album-level recognition (as in, telling you which CD you just plunked into your drive…in which case it usually cascades the track-level tags from the stored album data)
~fig
Is he doing it a whole album at a time? If so, it could simply check a database for albums whose length is X and whose songs lengths are y1, y2, etc. I think those two combinations alone can almost uniquely identify an album.
Broadcast Data Systems (BDS) also experimenting with a way of dealing with royalty payments that had to do with a pattern-recognition system to recognize songs played on the radio and various other media outlets as a way of figuring out airplay in order to calculate performance royalties. It’s used by Billboard for compiling weekly charts too.
I’m not sure if they add some kind of digital fingerprint or if they just scan the music to find easily recognized, unique patterns. If the latter, it would wokr with older vinyl if the songs are on file with BDS.
ETA: Here’s a bit more info on it. I’m not sure how helpful it is.
Gracenote also relies on users to submit CD & track data.
What do you know, thread on my research topic.
This is a very difficult problem, and it’s difficult to give a good answer since we don’t have complete information. How well known and popular is the music on the Vinyl? I also find it hard to believe he’s going through all the effort to rip it without entering ANY sort of identifying information AT ALL on the tracks. Even when I’ve relied on software to detect files later, I’ve still at least put in some basic file information so I could identify it later.
Anyway, waveform recognition is a highly researched topic right now, particularly because of technology like what the cell phone ad advertises. The problem is, it’s a difficult problem, and it works well for them because they probably use a relatively small database of the current popular songs. That is, if you were to hold it up against the Top 40 station, it will probably be accurate, but if you hold it to a classic rock station, it probably won’t recognize anything. Of course, that’s all WAG, since I have never played with it, and I don’t know exactly how they went about implementing it.
As for the Vinyl, I’m curious as to how well known the tracks are. If they’re something that a large percentage of the population is likely to rip, it’s POSSIBLE there’s some form of waveform recognition. I think it’s more likely, that it uses simple calculations to gather basic simple information instead of literally trying to compare waveforms (which is extraordinarily difficult). Someone upthread mentioned song lengths, this could vary within a few seconds, depending on how it’s ripped, but could be useful information. They could also easily grab certain other simple information using FFT (Fast Fourier Transform, allows you to extract frequency information), to grab strong pitches/chords, dynamics, etc. Not to mention, since FFT is part of the MP3 compression anyway, they can problem extract a lot of this sort of information with minimal effort, and use it to find close matches in a database.
IOW, while the technology does exist, depending on exactly what he was ripping, it may be possible. Still, as others said, I think it’s FAR more likely that there’s fault in the “My brother swears this is true” statement and he’s not giving you the whole story. Entering some basic information about an album (as little as the artist and/or album name) is FAR, FAR more reliable than any sort of automatic recognition, particularly in a space as large and as diverse as music.
Once his vinyl albums are on his computer, they’re digital music files. How do you think we get digital versions of old records? As for checking the files and providing artist and title information–a pretty common and unremarkable feature in CD ripping, even though CDs can’t possibly come with enough digital signatures to be recognized by every single media player program in existence, especially since there’s no computer filesystem on a music CD–I’ve always figured it was something simple like a checksum–even simpler if, like Blaster Master supposes, your brother is entering some of the information himself. If you know the name of the album, you can find album covers, artist and song title information on Google in all of 5 seconds.
If you’re ripping from a CD, recognition is almost always based on the CD’s TOC (table of contents), which is an encoded list of the exact starting and ending points (timewise) of each track. Based on that information it is usually possible to get a solid match, provided that your CD has already been entered into the database being queried.
Blaster Master, my company has been using track-level acoustic fingerprinting for recognition for several years now, and I believe Gracenote uses a similar technology in Itunes.
To my knowledge, there is nothing extraordinary about the OP’s claims of track recognition from ripped LPs. Once they’re digitized, the individual track file fingerprints will be indistinguishable from digital equivalents in Gracenote’s database.
This has happened to me at least once. In fact I can tell you what album it was. It was one of the two discs for Genesis’ “Seconds Out” live album. I was completely surprised when the track listing popped up, especially since I wouldn’t really consider it a popular album. I figured I split the tracks close enough to the actual album track time that the software was able to recognize it.
My wife had something similar (and somewhat startling) happen to her a couple years ago. One of the kids had left a disc in the drive for a Mickey Mouse or Winnie the Pooh game that they were playing. When my wife turned on the computer, the CD autostarted but it started the “CD Jukebox” application instead of the game, and the software thought that the CD was was some hardcore punk CD with f-bombs in just about every song title. My wife came in and asked me “What have you been listening to?” and I went in and popped open the CD tray and out came the Disney game CD.
The story of Joyce Hatto is interesting. A large number of recordings were published by her husband’s label late in her life, and he eventually admitted these were fraudulent copies of other pianists. While there had been questions asked about this otherwise unknown pianist having such a huge unpublished catalogue, it was Gracenote that made the killer blow, by identifying one of her discs as the original from which it was copied.
Yeah - classical music’s one real tabloid story of 2007. I’m still not sure her husband has come clean entirely about what Hatto knew during the process. I find it hard to believe that she wasn’t in on it.
I don’t know. There’s hundreds of people manipulating little old ladies for financial gain every day. The only difference here is that it wasn’t her who parted with the money.
You may be right, but the recordings came out while she was alive, and she presumably heard at least some of them. If she was as handicapped by her poor health as her husband claims (that was his given reason for stealing parts of other recordings in the first place), then she would have recognized the released recordings as not being what she produced in the studio. That is, unless she suffered from a severe lack of self-awareness about her playing. But there’s certainly a chance she was innocent. She may have just been another dupe in the process.
I was thinking that dementia could come into it?
Okay, I tried dragging all of my own recordings, (cassette tape and vinyl) into Itunes, and it did nothing at all with them 
When I tried the ‘get info’ option, it said that it couldn’t get info on them because they hadn’t been ‘imported with itunes’ (which I guess means ripped with itunes.)
Any idea how I can get the benefit of this amazing music recognition software without having a cell phone - or paying too much, ideally? 