To be more clear: While we cannot measure the original signal independently of
the noise, the processed signal's frequency content can be determined (I am
assuming voice), and if it is lacking in highs compared to an average human
voice then the processing could be made to crank up the highs. I think the
algorithm could automatically determine the general pitch of the voice (such as
male vs female) and vary the adjustment accordingly. Of course this would not
fit every voice with a given general pitch, and the user could be given various
settings with different amounts of high emphasis, that would be parametric
based on the general pitch (not like the existing equalizer).
73,
Erik K7TV
----- Original Message -----
From: Erik N Basilier
To: [email protected]
Sent: Thursday, November 20, 2008 9:29 PM
Subject: Re: [Elecraft] K3 NR
I find the K3 NR to be rather good, but what strikes me the most is how I
can't find a favorite setting, but I always have to go through the available
settings to find something that suits a given signal. That would suggest that
even more settings could be useful, if they bring anything new to the table. I
have to say that I have been rather impressed listening to sample files from
BHI LTD:
http://www.radio.bhinstrumentation.co.uk/html/sound_file_demonstration.html
If I am not mistaken, I think some time ago I saw some of their promotional
materials boasting that they used an algorithm based on neural networks. It
would be interesting to know if such an approach can be shown to be equivalent
to the approach used by Elecraft, or if it offers a useful variation.
Another thought: Many noise reducers including the K3's do seem to attenuate
the high frequencies, sometimes just slightly and sometimes excessively. Much
of what we perceive as irritating noise seems to be in the higher frequencies,
and just cutting those off using bandpass or equalizer might have a similar
effect. To make it easier to determine whether a particular NR setting is
useful, it might be helpful if each of the settings would include filtering
that would emphasize the highs to compensate for the highs lost in NR. I am
guessing that the effect of the NR would then appear more subtle, but also more
honest. Of course the design parameters might not easily translate to the
effect of audio frequency response, and that response may vary with the
momentaneous characteristics of the wanted signal and the noise, but at least
it should be possible to measure the signal and noise characteristics in real
time.
73,
Erik K7TV
------------------------------------------------------------------------------
_______________________________________________
Elecraft mailing list
Post to: [email protected]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com
_______________________________________________
Elecraft mailing list
Post to: [email protected]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com
