Path: utzoo!utgpu!watmath!clyde!att!osu-cis!tut.cis.ohio-state.edu!mailrus!nrl-cmf!cmcl2!polyof!kepler1!rjfrey
From: rjfrey@kepler1.UUCP (Robert J Frey)
Newsgroups: comp.sys.next
Subject: Re: Fwd: LSC problem
Keywords: FFT, noise suppression
Message-ID: <14@kepler1.UUCP>
Date: 17 Nov 88 18:21:08 GMT
References: <6843@batcomputer.tn.cornell.edu> <77955@sun.uucp>
Reply-To: rjfrey@kepler1.UUCP (Robert J Frey)
Organization: Kepler Financial Mgmt, Setauket, NY
Lines: 35

Regarding the notion of using an FFT to analyze white noise then cancelling
it out to increase signal-to-noise, trying to net out one white noise signal
with another will probably just mean more noise.  I have seen systems that
do what's being described here.  There's a two-part mike, one part is directed
outward so that it picks up ambient noise, the other is directed inward so
that it picks up voice plus ambient noise.  A simple circuit then nets out
the two signals.

If you wanted to get a little fancy you could perform a spectral analysis on
the two inputs separately, then try and analyze the difference.  The important
thing to remember is that white noise is by definition unpredictable.  If you
want to perform this sort of enhancement you need *simultaneous* samples.

Of course, there are lots of signal processing techniques that work on a
different basis: that of assuming there is something fundamentally different
between the character of the signal and the "noise".  For example, if the
signal-to-noise ratio is fairly low, then filtering out frequencies below
some threshold (amplitude or frequency) works okay. These techniques might 
have trouble filtering out alot of aperiodic noise which was "well mixed"
with the signal, and none of them attempt any real prediction of the noise
per se.  Some remarkable enhancement is possible, however.

If the noise were not "white", then an appropriate analysis might have a
chance at filtering it out. For example, certain over-the-horizon radars
employed by the USSR can disrupt shortwave broadcasts by overlaying an 8
Hz "clicking".  This type of interference might be amendable to a predictive
filtering.

==============================================================================
|Dr. Robert J. Frey               | {icus, spl1, dasys1}!acsm!kepler1!rjfrey |
|Kepler Financial Management, Ltd.|------------------------------------------|
|100 North Country Rd., Bldg. B   | The views expressed are wholly my own and|
|Setauket, NY  11766              | and do not reflect those of the Indepen- |
|(516) 689-6300 x.16              | dent Republic of Latvia.                 |
==============================================================================