[music-dsp] Killer speech noise reduction?

Peter Maas MPEG4_SA at hotmail.com
Mon Jun 24 16:53:11 EDT 2002


What about pattern recognition?
I've been spending a lot of time of investigating self-simularity
principles, noise isn't supposed to be simular at al...
in that case (and I'm online brainstorming here) it should be possible to
look for reoccuring spectral patterns.
Those are probably the speech parts....
Then you could gate the spectral envelope... just throw out the low mag
stuff...

You could also look for rapid changing frequencies... wasn't the technique
called cepstrum?
a voice signal would last for a couple of FFT frames... noise wouldn't...

(just some ideas...)

P

----- Original Message -----
From: "James Chandler Jr." <jchandjr at bellsouth.net>
To: <music-dsp at shoko.calarts.edu>
Sent: Monday, June 24, 2002 7:31 PM
Subject: [music-dsp] Killer speech noise reduction?


> A curiosity question--
>
> Does anyone know of programs that can enhance speech intelligibility that
is
> at or below the noise level?
>
> Have heard that FBI and CIA type organizations have really effective
speech
> extraction software, but maybe that is just rumor.
>
> Lawyers occasionally bring me "really important" evidence microcassette
> tapes that are hopelessly buried in noise. Said lawyers are routinely
> saddened and alarmed when I can't recover their important evidence (though
> these guys have not learned their lesson yet and made serious efforts to
> improve their surveillance recording techniques (GRIN)).
>
> Have used several common spectral noise reduction programs. "Noise
signature
> based" spectral noise reduction works fabulous if the signal is at least a
> little higher than the noise level.
>
> But commercial spectral noise reduction programs seem completely
ineffective
> as soon as the noise is as loud as the signal level. Once the S/N ratio
gets
> really bad, spectral noise reduction seems to make speech intelligibility
> worse rather than better. Not only does the speech remain unintellible,
but
> it also sounds like a robotic ring-modulated special effect.
>
> Are any specialized noise-reduction programs available, customized for
> improving speech intelligibility on really noisy recordings?
>
> Thanks
>
> James Chandler Jr.
>
>
>
>
> dupswapdrop -- the music-dsp mailing list and website: subscription info,
> FAQ, source code archive, list archive, book reviews, dsp links
> http://shoko.calarts.edu/musicdsp/
>

dupswapdrop -- the music-dsp mailing list and website: subscription info,
FAQ, source code archive, list archive, book reviews, dsp links
http://shoko.calarts.edu/musicdsp/




More information about the music-dsp mailing list