You should the software RX by Izotope, it will have a coloured spectrum analyser and you should be able to pinpoint exactly where the human vocals are and isolate them from the music and then save it, it should give you a 99% perfect vocal without any of the beat, although you may loose a bit of the quality of the vocals, this is the technique used by television stations when they record an "in the field" segment of the news where they are outdoors & need to remove background sounds like traffic, background noise ect.