If I recall correctly stuttering is often caused by some kind of additional delay in the perception of sound by the speaker. Anyone can be made to stutter by feeding their voice to their ears phase shifted. In Salim's case, he is probably trying to introduce a compensating delay, to reduce the phase shift in the voice feedback loop.

This sounds like a great project for the group to help with. Given his low bandwidth requirements, it should be pretty easy to implement delays, but how do we do pitch shift without changing tempo ?.

Alternatively, an ISD speech recorder chip might be coaxed to do the job, with a uC to handle its inputs.

Steve