Dan Luke writes:

It removes a lot of the static noise. I do not know the technical details
of how it does it. However, the freqencies of human speech do not cover the
audible spectrum, and all extraneous frequencies may be filtered quite
easily. Even frequencies at the upper and lower ends of human speech may be
filtered with minimal effect on intelligibility. Perhaps it's as simple as
that.

A noise-reduction system can make some assumptions about what is
_probably_ noise and what is _probably_ signal, but in most cases it
will drop at least part of the signal.

You need at least about 4 KHz for speech, but some sibilants have
components that go as high as 8 KHz or so, and if you chop off the
high frequencies the sibilants may be lost. Thus, 'f' and 's' might
start to sound the same, because the difference between them is in the
high frequencies.

Using fixed phraseology helps a lot, because it is highly redundant.
However, there is still the potential for confusion in relatively
random, non-redundant communications, such as strings of digits or
letters.

--
Transpose mxsmanic and gmail to reach me by e-mail.