Re: Noise source processing

-----Original Message-----
From: mgraffam@mhv.net <mgraffam@mhv.net>
Date: Thursday, August 06, 1998 10:17 AM


[[...]>The noise coming off of the sound card is more beige than white
though..
>
>Does anyone know of any papers, articles or whatever on good techniques to
>remove bias from the audio source?
>
>My initial thoughts are along the lines of just hashing everything, but
>this will be slow, and I'd like to see what other ideas are out there.

What about an analog filter placed before the digitizer? Nowadays there
should be inexpensive single-chip implementations of multiple-tap equalizers
made for consumer audio. You just have to straighten up the amplitude
ignoring the phase rotation, because the power spectrum is a real-codomain
function (being the Fourier Transform of the autocorrelation, which is an
even function). It won't be perfect, but it'll help.

Enzo

At 11:34 PM 8/6/98 +0800, Enzo Michelangeli wrote:

>[[...]>The noise coming off of the sound card is more beige than white
>though..
>>
>>Does anyone know of any papers, articles or whatever on good techniques to
>>remove bias from the audio source?
>>
>>My initial thoughts are along the lines of just hashing everything, but
>>this will be slow, and I'd like to see what other ideas are out there.
>
>What about an analog filter placed before the digitizer?

The engineering issue is this: no matter how great you filter out 
the junk, you don't have a perfect analog measurement to start with.
(Note that filtering out junk reduces info content).

So the question is how to distill the bits out of your binary digits.
You need to reduce the number of bits; you need to make them equiprobable.

I have not had luck*time implementing Shannon's method for distilling bits,
mentioned as suggestion #2 in the RFC.  This method produces perfect 0:1 ratio
but takes indeterminate input bits to do so.  Please post if you play with
this.





honig@alum.mit.edu