
Posted by John on 3/12/2006, 3:29 pm, in reply to "melody vs. harmony" Beyond that you would need to divide your recorded piece of music into several samples, like into units of .5 seconds. With each of those time units you could then produce a limited frequency spectrum, which would give you the frequency content of your recorded piece in snap shots of every .5 seconds. You would also have to filter the results based on relative amplitude to compensate for noise and harmonics, and then estimate the pitches based on the obtained frequency content. At this point you would need to then figure out which pitches are most likely the melody, which is another adventure in itself.
Link: http://jvalentino2.tripod.com/
70.246.63.14
You will first need to learn how to extract a frequency spectrum from a sampled sound, see http://jvalentino2.tripod.com/dft/index.html
Message Thread:
![]()
« Back to thread