Filtering base sound

Here is the scenario, I have a room with a known noise source, and a normal
sound source like a person talking, background noise etc.  If I record the
known source and also record the regular sounds can I use the recording of
the known source to filter itself out of the regular sound track?  Picture a
room with radio turned on but off station, white noise, and then having a
person speaking.  I need to record the person with a standard mike, not one
that I can put close to him/her but the background noise of the radio is
showing up on recording.  My thought was to somehow sync two recorders
together with a common time source and then digitally remove the radio from
the voice track.  Any thoughts?

Bill

Bill wrote:

> Here is the scenario, I have a room with a known noise source, and a normal
> sound source like a person talking, background noise etc.  If I record the
> known source and also record the regular sounds can I use the recording of
> the known source to filter itself out of the regular sound track?  Picture a
> room with radio turned on but off station, white noise, and then having a
> person speaking.  I need to record the person with a standard mike, not one
> that I can put close to him/her but the background noise of the radio is
> showing up on recording.  My thought was to somehow sync two recorders
> together with a common time source and then digitally remove the radio from
> the voice track.  Any thoughts?
> 

You can get some improvement but don't expect miracles to happen. There 
are several techniques that can be used (wiener filtering, spectral 
substraction, adaptive processing of the signal from several mikes). All 
of that helps in some extent however if the signal is severely garbled 
with the noise it is not going to be much help.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com

It almost certainly won't work - because of phase and amplitude
miss-matching. The adaptive filter goes a long way in trying to match
the phase and magnitude up again but even it is not perfect.

Naebad

the human ear/brain combination is hard to beat....

record the room sounds with two mics seperatee by about 1 foot, i.e
make a stereo or binaural recording, and then play back in  stereo into
stereo headphones...

if done correctly, the listener will be able to listen to the playback
and  "tune into" the desired voice .... otherwise known as the
"cocktail party effect" just as if  they were standing there in the
room


Mark

Actually I'm recording something that eventually will be used in a voice
processing situation, ie: Generating a transcript from the recording, so
while it may be possible for the human brain to 'filter out' the noise I'm
not sure even a high end voice recognition program could handle this thats
why I'm trying to figure out if I can filter using a base sound.

Bill

"Mark" <makolber@yahoo.com> wrote in message
news:1137127272.510823.198230@g44g2000cwa.googlegroups.com...
> the human ear/brain combination is hard to beat....
>
> record the room sounds with two mics seperatee by about 1 foot, i.e
> make a stereo or binaural recording, and then play back in  stereo into
> stereo headphones...
>
> if done correctly, the listener will be able to listen to the playback
> and  "tune into" the desired voice .... otherwise known as the
> "cocktail party effect" just as if  they were standing there in the
> room
>
>
> Mark
>

Mark wrote:
> the human ear/brain combination is hard to beat....
> 
> record the room sounds with two mics seperatee by about 1 foot, i.e
> make a stereo or binaural recording, and then play back in  stereo into
> stereo headphones...
> 
> if done correctly, the listener will be able to listen to the playback
> and  "tune into" the desired voice .... otherwise known as the
> "cocktail party effect" just as if  they were standing there in the
> room

I've used this technique to receive weak AM transmissions. Feeding the 
upper sideband to one ear and the lower to the other decorrelates the 
noise somewhat and improves the effective SNR by almost 3 dB.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:


> 
> I've used this technique to receive weak AM transmissions. Feeding the 
> upper sideband to one ear and the lower to the other decorrelates the 
> noise somewhat and improves the effective SNR by almost 3 dB.
> 

Oh yes. The selective fading make it sound very cool also. It creates 
nice stereo effects.

Vladimir Vassilevsky

DSP and Mixed Signal Design Consultant

http://www.abvolt.com