Newbee wants to kick start - second try

[Ian Stirling]

ramsesoriginal wrote:
> If it's like that, then a simple amplitude-modulation over a
> pre-recorded, fixed, carrier of the encrypted data should be no
> problem, since amplitudes are sent...

It's not.
It sends what is in essence a model of what state the human throat is in.
This is what the codec is designed to extract. It takes the audio 
signal, and extracts features that could have been generated by a human 
throat/mouth, and throws away the bits it thinks the human won't notice.

It then encodes this information in such a manner that small bit errors 
cause minor sound degradation on the other end.
On the other end, a reverse process is performed.

The fact that any non-vocal signal at all can get through is a 
byproduct, and the system is _not_ robust to this.