“I want to play a CD and be able to chart the music to it. Is there anything out there that can help me do this? Thanks, James”
Before I get into the nuts and bolts at all I’d like to give a quick answer to this question: Yes. Try the Library of Congress or contact the publisher(s) of the songs on the CD. If the music’s been registered as a copyright, then there could very well be a chart for it there (a chart used to be a required element to filing a copyright… no longer, however). Another option is one of the several new online notation services that will transcribe music you send to them – they’ll even take the audio file electronically if you have that means available to you. A broke local college student with music theory training could probably do the job also. That’s probably not going to be in the time frame or cost allowance you were hoping, but that’s the real solution at this point. Now on to the entertainment portion of our program, and more about what you were probably actually wanting to know.
Wow – every once in a while a question comes along that goes beyond just technical mumbo-jumbo, and takes us to the area of “what if…” The reality here, of course is that what you’re asking for is genuine and cognizant music recognition by a computer; an ability to discern, as reliably as a human can, the timbres and tones of contemporary and classical instruments. What’s important to understand here is that the manner in which you, a consciously thinking human, can play a jazz CD and easily tell which instrument is the piano, which is the upright bass, the saxophone, the drums, and so on, is just a bit beyond what our machines can accomplish right now. There are a few applications in existence (mostly for PCs) that can create notation from a recording of a single polyphonic instrument (such as piano or guitar), but even those have to be told what instrument to expect. That is, they don’t just know that it’s a piano or a guitar, which is a very important element of music recognition. So, sorry James, the best we can do right now is one lonely, unidentified instrument at a time. Spectrally analyzing the recorded audio of any kind of ensemble and extracting the relative note values is one thing, but being able to subdivide that information into the different instruments of an ensemble is a whole lot more difficult – getting into the territory of artificial intelligence difficult! What you’d be asking a computer to do in this case is ‘listen’ to a song, be able to differentiate between instruments (especially those that have similar timbres and/or overlapping frequency ranges), and then convert all of that information first into MIDI, and then into a multi-staved piece of correctly notated music. I’m afraid shortly after we achieve that pinnacle of technology we might encounter the machine that also critiques music – play a computer music and it tells you how bad it thinks it is! Hope they can program them without the typical elitist art squawker obfuscation you might normally encounter in a music critic! Hey, I could do that, couldn’t I?