We have a DSP which I assume could be programmed to remove vocals. Has anyone done any work in this area or given this any thought?

Vocal removal isn't one of the available functions of the unit's DSP. Even if it were, you have to remember that vocal-removal is iffy at best. Doesn't work on all tracks, and on the ones where it does work, the remaining track doesn't sound quite right. Proper karaoke would be with a proper music-only track.

Still, it would be fun to have a vocal-removal algorithm to play with. Doesn't have to be in the DSP, it could be in the kernel. It would be implemented in much the same way as Voldaj is currently.

And I like the name: kaRIOke. That's cute.