Yes this is basically what most commercial noise-reduction software does [i=
n a very general sense]. It analyzes a sample of the 'noise' then applies =
a 'fix' [based on that] to whatever longer segment of the recording you are=
trying to fix.
--- In "jmccubbinmd" <mccubbin.astronomy@=
...> wrote:
>
> I'm relatively new to recording outdoor ambient sounds, but am very intri=
gued by it. I have done plenty of audio recording (amateur), but also have=
a background in astroimaging. I have been trying to solve the noise floor=
problem with microphones, as it seems to bug everyone to some degree. It =
just seems difficult and expensive to get rid of it. So that got me thinki=
ng.
>
> For you audio engineers out there, please tell me if this is solid thinki=
ng.
>
> When I take an astrophotograph, our exposure times are very long. The ch=
ip suffers from electronic noise, not unlike mic noise. To get rid of it, =
we "calibrate" our images. One of the things we do is to take a "dark fram=
e" of equal length, at the same temperature, and same degree of thermoelect=
ric cooling of the chip. Then we mathematically subtract the noise from th=
e image and the image improves dramatically.
>
> As I was cycling through the countryside today it struck me that might be=
possible with microphone noise. So I tried it. I did some more recording=
of my test subject (for now), the pond at our farm. I got a clip of a bir=
d chirping (my wife thinks it's a warbler, I'm not an expert on bird calls,=
so and ID would be appreciated).
>
> I then recorded a 10 second clip with the microphones at the same gain in=
as quiet conditions as I could find in my study (full of books and very qu=
iet). I used that as my "noise floor sample".
>
> Then I imported the audio into Audacity (for simplicity) and first amplif=
ied the bird call sample a fixed amount (19db in this case). I then amplif=
ied the noise floor sample by the same amount. I then did a high pass filt=
er to get rid of a distant train passing (troublesome buggers).
>
> Then I copied the clip with and without the noise reduction. In audacity=
, you first take a "noise sample reference", then apply that reference to t=
he redording. Obviously, I used the noise sample and applied it to the rec=
ording of the bird. The first half of his mp3 is without noise reduction, =
just the 19db amplification and hipass filter. The second half is with noi=
se reduction.
>
> I think it worked.
>
> If there is a better way of doing this, let me know.
>
> The equipment used here is a "matched" pair of Rode NT1-A's with foam win=
d guards, into a Sound Devices 702. The gain was set at 61.8db both for th=
e recording and for the noise sample as I felt this needed to be exactly th=
e same for both.
>
> Here is the comparison sample:
>
> http://m57.net/downloads/sfp/warblertest.mp3
>
|