THe problem is, that sound tends to be recorded over many different frequencies. What you want is the different channels that the sounds is recorded, but that it is impossible unless you actually have the seperate recording tracks, which would be next to impossible. The only thing you could really do is use something like ACID and try to use the frequency mapper to select a range of frequencies in which the sound is recorded. But you'll also be extracting whatever sounds are recorded on that frequency also.