11 Sep
2016
11 Sep
'16
8:14 p.m.
Hi all,
Does anyone have access to an easy to implement (as in code)
dereverberation algorithm assuming one has access to the room impulse
response? I am looking for a way to clean-up recorded audio signal in
order to improve signal clarity (e.g. speech). Any info on this topic is
most appreciated.
Best,
--
Ivica Ico Bukvic, D.M.A.
Associate Professor
Creative Technologies in Music
ICAT Senior Fellow
Director -- DISIS, L2Ork
Virginia Tech
School of Performing Arts – 0141
Blacksburg, VA 24061
(540) 231-6139
www.performingarts.vt.edu
disis.music.vt.edu
l2ork.music.vt.edu
ico.bukvic.net
11 Sep
11 Sep
9:02 p.m.
Hi,
I wonder if a multiband gate/expander could help?
Iain
Em 11/09/2016 17:14, Ivica Ico Bukvic escreveu:
Hi all,
Does anyone have access to an easy to implement (as in code)
dereverberation algorithm assuming one has access to the room impulse
response? I am looking for a way to clean-up recorded audio signal in
order to improve signal clarity (e.g. speech). Any info on this topic
is most appreciated.
Best,
10:27 p.m.
Can you elaborate?
--
Ivica Ico Bukvic, D.M.A.
Associate Professor
Computer Music
ICAT Senior Fellow
Director -- DISIS, L2Ork
Virginia Tech
School of Performing Arts – 0141
Blacksburg, VA 24061
(540) 231-6139
ico(a)vt.edu
www.performingarts.vt.edu
disis.icat.vt.edu
l2ork.icat.vt.edu
ico.bukvic.net
On Sep 11, 2016 17:02, "Iain Mott" <mott(a)escuta.org> wrote:
Hi,
I wonder if a multiband gate/expander could help?
Iain
Em 11/09/2016 17:14, Ivica Ico Bukvic escreveu:
Hi all,
Does anyone have access to an easy to implement (as in code)
dereverberation algorithm assuming one has access to the room impulse
response? I am looking for a way to clean-up recorded audio signal in order
to improve signal clarity (e.g. speech). Any info on this topic is most
appreciated.
Best,
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
12 Sep
12 Sep
12:07 p.m.
On Sun, Sep 11, 2016 at 04:14:33PM -0400, Ivica Ico Bukvic wrote:
Does anyone have access to an easy to implement (as in
code)
dereverberation algorithm assuming one has access to the room
impulse response? I am looking for a way to clean-up recorded audio
signal in order to improve signal clarity (e.g. speech). Any info on
this topic is most appreciated.
This is not an easy matter. There is no single 'room impulse response',
there is a different one for every combination of source and receiver
position. And even these can't be assumed to be constant except maybe
at low and low mid frequencies.
So in practive any given room IR is just a hint, and algorithms
that work need to be adaptive. They usually work in the frequency
domain, and build up statistics of dynamics in each frequency band.
Using these it's possible do attenuate some of the reverb. How
much depends strongly on the source material. This is still a very
active research topic. Most research is related to speech pickup,
not music.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
12:17 p.m.
On Mon, 12 Sep 2016 12:07:10 +0000, Fons Adriaensen wrote:
How much depends strongly on the source material. This
is still a very
active research topic. Most research is related to speech pickup,
not music.
Sometimes, when I watch documentations with interviews, I guess the
easiest way to solve the issue, would have been to chose another
location for the interviews ;).
1:03 p.m.
Hey hey,
if you're not absolutely set on open source software, there is iZotope RX 5:
https://www.izotope.com/en/products/repair-and-edit/rx.html
Noty sure if it would run under Linux, but usually they offer 10 day trial
versions. So, if all else fails, it might be a possibility. They certainly
have dereverberation and the demos I heard sounded quite good.
Ta-ta,
----
Ffanci
* Homepage: https://freeshell.de/~silvain
* Twitter: http://twitter.com/ffanci_silvain
* GitHub: https://github.com/fsilvain
12:41 p.m.
Le dimanche 11 septembre 2016 à 16:14 -0400, Ivica Ico Bukvic a écrit :
Hi all,
Does anyone have access to an easy to implement (as in code)
dereverberation algorithm assuming one has access to the room
impulse
response? I am looking for a way to clean-up recorded audio signal
in
order to improve signal clarity (e.g. speech). Any info on this topic
is
most appreciated.
Best,
A transient designer can be used to effectively reduce the reverb.
Calf-plugins has one.
Guillaume
1:05 p.m.
On Mon, Sep 12, 2016 at 02:54:24PM +0200, Guillaume Meurisse wrote:
A transient designer can be used to effectively reduce
the reverb.
Which is basically an expander/gate that, instead of following the
input signal level with some attack/release timing, imposes a
predefined envelope once triggered.
Dereverberation algos do something similar in multiple (30..100)
frequency bands, with the parameters derived from the signal
statistics. They also will trigger on level changes rather
than absolute levels. This can get very complex.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
7:55 p.m.
Hi,
On Sun, 11 Sep 2016 16:14:33 -0400 Ivica Ico Bukvic <ico(a)vt.edu> wrote:
Does anyone have access to an easy to implement (as in
code)
dereverberation algorithm assuming one has access to the room impulse
response? I am looking for a way to clean-up recorded audio signal in
order to improve signal clarity (e.g. speech). Any info on this topic
is most appreciated.
If you have the impulse-reponse of the original room, then it should be
just a matter of applying the reverse of that via fft.
Incidentally that is just what drc (Digital Room Correction) [1] is
doing. Maybe try that?
- Arnold
[1] http://drc-fir.sourceforge.net/
9:07 p.m.
On 09/11/2016 10:14 PM, Ivica Ico Bukvic wrote:
Does anyone have access to an easy to implement (as in
code)
dereverberation algorithm assuming one has access to the room impulse
response? I am looking for a way to clean-up recorded audio signal in
order to improve signal clarity (e.g. speech). Any info on this topic is
most appreciated.
a common way to achieve dereverberation in speech processing is to use
spatially separated microphones (e.g., binaural), and use the
inter-microphone coherence function as an estimator of the gain. There
are different approaches of calculating the coherence function,
basically they are all frequency-domain based (but often with short
analysis blocks/low resolution), and require some temporal smoothing.
The artefacts are annoying if set to strong filtering, but subtle
settings may decrease listening effort and are (especially by hearing
impaired listeners) preferred over unprocessed signals.
Below I added are some references, you will see that the basic algorithm
was available already in the 70ies... If you need any of the
publications or some code hints let me know off-list.
Best,
Giso
@Article{Allen1977b,
Title = {Multimicrophone signal-processing
technique to remove room reverberation from speech signals},
Author = {Jont B. Allen and D. A. Berkley and J.
Blauert},
Journal = jasa,
Year = {1977},
Month = oct,
Number = {4},
Pages = {912-915},
Volume = {62}
}
@Article{Kollmeier1993,
Title = {Binaural Noise-reduction Hearing Aid
Scheme with R
eal-time Processing in the Frequency Domain},
Author = {B. Kollmeier and J. Peissig and V. Hohmann},
Journal = {Scand Audiol},
Year = {1993},
Pages = {28-38},
Volume = {Suppl. 38}
}
@Article{Wittkop2003,
Title = {Strategy-selective noise reduction for
binaural digital hearing aids},
Author = {Wittkop, Th. and Hohmann, V.},
Journal = {Speech Communication},
Year = {2003},
Pages = {111-138},
Volume = {39}
}
@Article{Grimm2009a,
Title = {Increase and subjective evaluation of
feedback stability in hearing aids by a binaural coherence based noise
reduction scheme},
Author = {Giso Grimm and Volker Hohmann and Birger
Kollmeier},
Journal = {IEEE Transactions on Audio, Speech and
Language Processing},
Year = {2009},
Month = {September},
Number = {7},
Pages = {1408-1419},
Volume = {17},
Doi = {10.1109/TASL.2009.2020531}
}
14 Sep
14 Sep
4:46 p.m.
Thank you, Giso, for sharing this.
Given in my case I am not concerned with mangling the audio (it is not
intended to be played back, only used to calculate TDOA between mics), I
have no reservations towards resorting to destructive kinds of
processing as long as it improves my ability to identify accurate TDOA
between multiple mics.
Best,
Ico
On 9/12/2016 5:07 PM, Giso Grimm wrote:
On 09/11/2016 10:14 PM, Ivica Ico Bukvic wrote:
Does anyone have access to an easy to implement
(as in code)
dereverberation algorithm assuming one has access to the room impulse
response? I am looking for a way to clean-up recorded audio signal in
order to improve signal clarity (e.g. speech). Any info on this topic is
most appreciated.
a common way to achieve dereverberation in speech processing is
to use
spatially separated microphones (e.g., binaural), and use the
inter-microphone coherence function as an estimator of the gain. There
are different approaches of calculating the coherence function,
basically they are all frequency-domain based (but often with short
analysis blocks/low resolution), and require some temporal smoothing.
The artefacts are annoying if set to strong filtering, but subtle
settings may decrease listening effort and are (especially by hearing
impaired listeners) preferred over unprocessed signals.
Below I added are some references, you will see that the basic algorithm
was available already in the 70ies... If you need any of the
publications or some code hints let me know off-list.
Best,
Giso
@Article{Allen1977b,
Title = {Multimicrophone signal-processing
technique to remove room reverberation from speech signals},
Author = {Jont B. Allen and D. A. Berkley and J.
Blauert},
Journal = jasa,
Year = {1977},
Month = oct,
Number = {4},
Pages = {912-915},
Volume = {62}
}
@Article{Kollmeier1993,
Title = {Binaural Noise-reduction Hearing Aid
Scheme with R
eal-time Processing in the Frequency Domain},
Author = {B. Kollmeier and J. Peissig and V. Hohmann},
Journal = {Scand Audiol},
Year = {1993},
Pages = {28-38},
Volume = {Suppl. 38}
}
@Article{Wittkop2003,
Title = {Strategy-selective noise reduction for
binaural digital hearing aids},
Author = {Wittkop, Th. and Hohmann, V.},
Journal = {Speech Communication},
Year = {2003},
Pages = {111-138},
Volume = {39}
}
@Article{Grimm2009a,
Title = {Increase and subjective evaluation of
feedback stability in hearing aids by a binaural coherence based noise
reduction scheme},
Author = {Giso Grimm and Volker Hohmann and Birger
Kollmeier},
Journal = {IEEE Transactions on Audio, Speech and
Language Processing},
Year = {2009},
Month = {September},
Number = {7},
Pages = {1408-1419},
Volume = {17},
Doi = {10.1109/TASL.2009.2020531}
}
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user(a)lists.linuxaudio.org
http://lists.linuxaudio.org/listinfo/linux-audio-user
5:33 p.m.
On Wed, Sep 14, 2016 at 12:46:34PM -0400, Ivica Ico Bukvic wrote:
Given in my case I am not concerned with mangling the
audio (it is
not intended to be played back, only used to calculate TDOA between
mics), I have no reservations towards resorting to destructive kinds
of processing as long as it improves my ability to identify accurate
TDOA between multiple mics.
But earlier:
>>Does anyone have access to an easy to
implement (as in code)
>>dereverberation algorithm assuming one has access to the room impulse
>>response? I am looking for a way to clean-up recorded audio signal in
>>order to improve signal clarity (e.g. speech).
So now I'm a bit confused... Do you need improved audio, or just
signal statistics ???
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
5:45 p.m.
On 9/14/2016 1:33 PM, Fons Adriaensen wrote:
On Wed, Sep 14, 2016 at 12:46:34PM -0400, Ivica Ico
Bukvic wrote:
Good point. As long as the spectral content is more accurate because it
is devoid of extraneous reflections and therefore potentially altering
the spectral composition to the point it fails to be cross-correlated
correctly, and a single window of sound is enough to do a more accurate
cross-correlation, I am fine with that. The in-between-windowing
artifacts do not concern me as long as they are not a cause of signal
degradation to the point where cross-correlation is affected by it. Hope
this helps. Thank you all for your input so far.
Best,
Ico
Given in my case I am not concerned with mangling
the audio (it is
not intended to be played back, only used to calculate TDOA between
mics), I have no reservations towards resorting to destructive kinds
of processing as long as it improves my ability to identify accurate
TDOA between multiple mics.
But earlier:
>> Does anyone have access to an easy to
implement (as in code)
>> dereverberation algorithm assuming one has access to the room impulse
>> response? I am looking for a way to clean-up recorded audio signal in
>> order to improve signal clarity (e.g. speech).
So now I'm a bit
confused... Do you need improved audio, or just
signal statistics ???
Ciao,
2786
days inactive
2789
days old
linux-audio-user@lists.linuxaudio.org
12 comments
9 participants
participants (9)
-
Arnold Krille -
Ffanci Silvain -
Fons Adriaensen -
Giso Grimm -
Guillaume Meurisse -
Iain Mott -
Ivica Bukvic -
Ivica Ico Bukvic
-
Ralf Mardorf