Subject: Re: [linux-audio-user] Noise removal, what filter is best?
From: RTaylor (ricktaylor_AT_speakeasy.net)
Date: Tue Aug 26 2003 - 13:26:39 EEST
On Tue, 26 Aug 2003 08:38:07 +0100 Anahata <anahata_AT_treewind.co.uk> wrote:
I have a bunch of files like that...
I used a demo denoiser that I mistook as freeware... and stupidly deleted
the originals before I listened to the real output of the program.
The only way to really remove them is to cut them out. Seeing as they're
evenly placed... You can create some nifty sound samples by altering the
pitch a bit and putting a delay on them. Some of it sounds pretty gorgeous.
The best thing I've found for noisecarving is a program called in-tune
{Australian... I believe it's been competed away by some commercial
program called intune} It's windows like d/noise. If you can find it {I
can't} it seems like it might run under wine. {It may have even come with
source} Seems to me a lot of small windows sound utilities might.}
I have had good luck at cleaning stuff up with just the basic tools in
snd {{reverb dontcha' know...} :} You can twist the hell out of it too.}
You might try some of the FFT stuff or even get fancy like this:
http://ccrma-www.stanford.edu/guides/planetccrma/node48.html
...Maybe try mixing it with another file...
> sqeak. I'm surprised that gramofile does not do anything for it, but I
> think your best bet in this case may be to use a sound editing tool like
> audacity and manually cut the noise blips. If they are much higher
> amplitude than the normal sound they should show up on the display;
> otherwise you'll just have to locate them by ear. When you find one you
> can either cut the noise out (shortening the recording by the length of
> the section you've cut out) or use the amplitude envelope tool to make a
> rapid fade out/fade in over the noise. The first method will cause a
> time discontinuity: the second may cause an audible droput in the sound.
> It depends on the nature of the recorded material and the length of the
> noise burst and you'll have to experiment to see which is the least
> objectionable.
>
> If the noise is at exactly one frequency, it may be possible to select
> a short section containing the noise and apply a filter with audacity,
> with a deep notch at the frequency of the noise. I wouldn't advise
> applying such a filter all through the recording because it would
> probably not sound very nice, and it's a pity to spoil a minute or good
> sound just to reduce the effect of a fraction of a second of noise.
> Again you'd have to do this for each noise burst, but with a bit of luck
> (i.e. if the frequency is the same each time) the same filter setting
} will work on each one.
This archive was generated by hypermail 2b28 : Tue Aug 26 2003 - 13:26:55 EEST