You need some kind of plugin to visualize each track (I use the one built-in to ReNoise, but I'm sure there are plugin VSTs if your DAW doesn't come with anything). Then, you can clearly SEE which frequencies each voice is hitting. From there, the task is to find out where the tracks overlap, and then decide which one gets EQ'd. For example, the high voice might have a lot of bass in it, but since it's the high voice you can (usually) just cut the low end right off it. Vice versa with the bass.
Another technique, which may also be helpful if you don't have any visualization plugins, is to make an EQ with a PEAK on a track; you "sweep" the peak back and forth while the song plays. As you sweep the peak across the frequencies, you listen for when the track becomes even MUDDIER/MESSIER, and then you know to replace that peak with a CUT in that frequency range on that track.