Newsgroups: comp.ai.neural-nets
From: jimmy@ecowar.demon.co.uk (Jimmy Shadbolt)
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!europa.eng.gtefsd.com!howland.reston.ans.net!agate!hilbert.dnai.com!redstone.interpath.net!news.sprintlink.net!demon!ecowar.demon.co.uk!jimmy
Subject: Re: Q: Selective attention for time series 
Distribution: world
References: <Cv3Ln1.GJq@ie.utoronto.ca>
Organization: Econostat
Reply-To: jimmy@ecowar.demon.co.uk
X-Newsreader: Simple NEWS 1.90 (ka9q DIS 1.21)
Lines: 34
Date: Wed, 31 Aug 1994 10:19:42 +0000
Message-ID: <778328382snz@ecowar.demon.co.uk>
Sender: usenet@demon.co.uk

In article <Cv3Ln1.GJq@ie.utoronto.ca> Thas@orion.rose.utoronto.ca writes:

>
>Greetings everyone,
>
>I am currently evaluating the use of Neural Nets (mainly single layer backprop)
> 
>as adaptive noise cancellation engines as illustrated below.
>Problem description:
>
>Noise reference ---------------
>                              |
>                              |
>                              Y
>       (Neural Net-that predicts the noise component)
>                              |
>                              |
>                              Y
>Corrupted signal --------->(Subtract)----->noise free signal (hopefully)
>(signal+noise)                               
>

This is similar to a "noise" embedding for signal detection described
by Jaroslav Stark (University College London) a few years ago in Int.
Journal of Bifurcation and Chaos (sorry I can't recall the exact title -
I took his course some time ago).
 
However, I remember that you can extract signals attenuated by as much as
-80 dB if you don't know exact form and even more than -100 dB if you do ..

Good luck

Drago

