Re: [HCDX]: sound processing results ?

[Vittorio De Tomasi <vdetomasi@xxxxxxxxxxxxx>]

Hi Guido,

I did several experiments with homemade DSP systems and software to
clean out DX broadcastings, with controversial results. The main reason
for this is that the signal that one wants to denoise is quite difficult
to describe: there is speech, music, and so on. The mathematical model
of the signal on which denoising science (or art ?!?) is based is not so
sophisticated, and tries to describe the frequency spectrum of incoming
signal with a short polynomial approximation. The story is quite
different if you want to process a CW signal: in that case you get
easily very good results, because the signal model fits very well the
desired signal!

I usually prefer to apply a properly cut band pass filter to the signal
I receive: the sharp edges you can get from a DSP filter help a lot to
remove noise without losing signal, and usually this is better than an
adaptive denoiser.

Adaptive denoisers (LMS filters) work reasonably well on voice signals,
but they need enough signal level to work, otherwise the filter will not
be able to track the incoming signal. This denoiser works quite better
if a good AGC is placed in front of it (an audio AGC, not the RF one).

I tried also frequency clipping denoisers (so called "NIR" filters, a
poor implementation is available in CoolEdit), but they are inferior to
adaptive filters, because they add too much artifacts to the signal
(this is an obvious consequence of the way they work).

Unfortunately DSP denoisers can do little, when applied after the AM
detector....

vy 73

Vittorio


Guido Schotmans wrote:
> 
> I have been trying some DSP and sound processing tools on a recording from
> Radio Vanuatu I made in early December.  It was a very noisy recording from
> which I could understand nothing.  I've ID'ed them only from the 'yellow
> bird call'.  I thought that it would be possible to do some sound processing

< snip >
-- 
*************************************************************************
Vittorio De Tomasi         ik2czl@xxxxxxxxx
Home page:                 http://space.tin.it/scienza/vdetomas 
My DSP page:               http://www.freeyellow.com/members/padan

"You dialed an imaginary number. 
Please rotate your telephone of 90 degrees and redial".
------------------------------------------------------------------------
This is a message from Vittorio De Tomasi <vdetomasi@xxxxxxxxxxxxx>
to hard-core-dx@xxxxxxxxxxxxx list. To unsubscribe the list, send
"unsubscribe hard-core-dx" in mail body to majordomo@xxxxxxxxxxxxxx
For more information, please check http://www.kotalampi.com/hard-core-dx/
or email Risto Kotalampi, risto@xxxxxxxxxxxxxx
------------------------------------------------------------------------