You should try to somehow get in touch with Imogen Heap. She likes to do weird things with technology and collaborated with others to build gloves that work as a performance tool. Having something where she could use her voice to control other instruments sounds like it would be right up her alley.
Story
Actually, now that I think of it, you can do what I described with some high-end vocoders that have patch-out jacks for the individual bands. Of course, that sort of option doesn't exist with any sort of digital vocoding device. Things like the VO-1, or the EHX Voice Box are intended to be leave-the-driving-to-us solutions for those who simply want the effect and aren't planning any experimentation.
Vocoders are a bit like EQ units. "Better" equalizers usually have more bands, like 15 or 31, simply because the manufacturer doesn't know which frequencies you might need to reduce or accentuate, so they give you lots of fixed bands to choose from, under the assumption that some of those will be what you need. The alternative is parametric equalizers that allow the user to select the resonant band to cut or boost. Those will generally not have more than 4 bands of adjustment.
"Good" vocoders will have more bands. Eight (like the old PAiA vocoder) is passable, but 12 or more is preferred to have a more realistic voice-like quality, and digital units will mimic 24 or more bands. Like equalizers, those bands are fixed and chosen by the manufacturer, under the assumption that - whatever the nature of the voice or other audio input you want to modulate it with - those bands will yield a reasonable approximation of voice.
What I'm aiming for is more like a parametric EQ, where the user gets to select which bands of their voice input the circuit responds to. I'm thinking that one control-band would respond to overall voice amplitude, regardless of frequency, another would respond to anything over1khz or so, and two adjustable ones could be tuned to the range of one's choosing. So, for instance, one could utter a "ssssss" sound to drive the >1khz section, and gradually add more voice to drive the tuned sections. The critical thing is that everything has to be adjusted so that one can actually produce the critical bands of vocal energy at will. I highly doubt that anything like this would be able to make an audio signal "talk". For that, I have one of those Danelectro "Free Speech" pedals.
