Correct me.OK :-)
Functional explanation: A voiced/unvoiced detector checks whether there is some signal above a certain frequency, but no signal below this frequency. If both both conditions are given it activates its output.
Electrical explanation: A voiced unvoiced detector is basically a state variable filter, which splits the incoming signal into a hipass path and a lopass path. You may think of it like of a crossover in a loudspeaker.
In both pathes there is an gate-detector (envelope follower with comparator).
In the lopass path the gate-detector is followed by an inverter. Both pathes deliver a logical signal. Those signals are AND'ed. The output of this AND-gate is the output of the "module".
Usually a VCA or switch which activates a white noise signal is controlled by the v/uv-detector.
And to my humble opinion: this kind of unvoiced detection is not very useful. At least it should work like:
In the hipass path there should be only an envelope follower. The output of the gate detector in the lopass path should activate this envelope follower. This creates a bit more dynamic unvoiced hissing.
But to be honest: understandability-wise the best and at the same time the cheapest solution is simply: put a highpass filter at 6-8kHz on the original modulator signal (the original voice). Mix this highpass filtered signal to the output of the vocoder. Its perfekt!
Florian