The trouble with ripping audio is that you usually can't just rip the audio itself unless it follows a standard formatting. If it were that simple, you could just set a break on anything being written to the AI interrupts and copy the binary of said size at the address sent. The trouble is that you also need the code used to process the sound.
The solution so far used in the .usf format was to take a savestate and a rom, then strip both down so nothing is left but audio and the minimal code to process it. HCS runs that project, but at the moment his site is down. There should be a message when it gets up and running at
http://hcs.freeshell.org/ It is usually music-specific, but it can be made to work with sound effects as well, I assume.
So, really, what I'm saying is that it would be a serious bite to try to do it, and that even when you do manage to get it going you'll still need to rip the content and save it in a far more orthodox format.