sound filter software

[Admin]

Does anyone know of software capable of filtering out background noise?

I have some videos with a droning airconditioner in the background, another with static/beep sounds in the background, and others with wind sounds.

If I could perhraps see the waveform divided into different frequencies and just use a mouse to mute certain parts, that would work too.

[hgordon]

Try using a sound editor software like Audacity (it's free and cross-platform).  You should be able to use some of the filters (e.g. FFT, Equalization, etc) in the "Effect" menu to clean up the sound.

[Admin]

I just got around to trying out Audacity . . . yea, I know this post is like 3 years old . . . but still . . .

Anyway, it sounds better with the filter than without, but it alters the sound to where its like a robot talking . . . I also played around with both the low and high pass filters with no luck.

Its a loud stable background hissing sound, overlaid with a man talking.

(actually, a robot talking, but its just a recorded human voice)

Anyone know any other free software?

[corrado33]

I don't know any FREE software, but I know Soundtrack pro in apple's Final Cut Studio can do it.  Basically it takes a sample of "silence" and subtracts it from the rest of the sample.  Silence being everything you want to be removed (like the air conditioner).  So if you had a friend who had it, then you could do it that way...  I don't know if this would work for the beeps though...

EDIT:  My cousin is a videography major.  We record weddings and events sometimes, and he has the software.  Very powerful stuff if you ask me.  That's pretty much all I can do on it though.   

[Admin]

Its a 1 min long 3.6mb .wav, if anyone wants to have a go at it.

[corrado33]

We'll take a look at it if you'd like.  I'll see what I can do with it.  My e-mail is in my profile.  It's gmail so 3.6 megs is fine for attachments.

[Admin]

You'll find two files, the audio file, and my best attempt at removing the noise:
http://www.societyofrobots.com/uploads/sound/

That said, I just tried out the 1.3.12-beta version of Audacity. The noise reduction artifacts are greatly reduced compared to the v1.2, but still not good enough for me. My best attempt file uses this beta.

[corrado33]

I'll look at it tonight, I don't have head phones now, so I can't really do anything with it without being really obnoxious.  I can definitely do it though, as I can see I've made a difference to the noise level through the waveform.

[corrado33]

Hey Admin, I have my three best tries... They all sound similar.  The only "noise" that's left is a little bit of "garbling", but no matter what I do, I can't seem to get rid of it.  I've tried all the effects and filters that seemed relevant.  I don't know how you want me to send you the files though.  Let me know how you want them.  Ah n/m, check your private messages for a link to them...

Now back to writing my paper...

"For a LONG time, no one really..."  

I must have listened to that thing a couple hundred times.  I can pretty much recite from memory haha..  It almost seems that it would have been easier if you found a friend that sounded like this person, and got them to reread it with the exact same pacing etc...  Might take an hour or two, but at least you'd have good source audio.  It's like my cousin always says "It's better to fix something during production than try to do it post production." (Or something like that... it's better to try to fix something WHILE you're recording it, rather than say "We'll fix it in post")  

EDIT:  Well, I think I sent you a private message... I'm not really sure, if you don't have one from me, just let me know.

[Admin]

Hmmm it sounds like Audacity v1.2 and the software you used has the very same noise reduction algorithm. The 'speaking into a tunnel' sound artifact is pretty much the same for both. Its open source, so very possible.

The v1.3 beta no longer has that tunnel sound, but the noise is still audible, albeit much reduced.

And yea, I had no idea during production that there was a background static noise. A proper recording wasn't possible given the situation. Otherwise I'd just redo it.

I think I'll wait a bit longer to see if anyone has other ideas, then I'll just have to go with what I got . . .

[Soeren]

Hi,

Hmmm it sounds like Audacity v1.2 and the software you used has the very same noise reduction algorithm. The 'speaking into a tunnel' sound artifact is pretty much the same for both. Its open source, so very possible.

The v1.3 beta no longer has that tunnel sound, but the noise is still audible, albeit much reduced.

They must have changed the algorithms a bit, but if you filter too aggressively, any kind of DSP will do that.

I hauled it through CoolEditPro to analyze it and it's very little of the noise you can remove without touching the voice, as most of the noise is in the speech band.
If you had a take of the noise alone, but otherwise identical in timing and frequencies, it would be easy to subtract, but with a signal like yours, the best that you can hope for is cutting anything outside the passband, using notch filters sparingly (max. around 15..20 dB) and masking what's left, perhaps running it through a chorus box or similar.
In the old days of sliders and turners, it would be a lot faster, tuning the notches while listening, instead of writing a number, listening, changing the number slightly etc.

One other thing that may or may not help is to  morph the sound when cleaned up as much as possible, that might remove some more noise and the "vibrato" resulting from artifacts of the filtering.

Masking with soft background music is your next tool - the right kind of classical music can actually enhance the focus and attendance of the listener.

If the exact voice doesn't matter, besides morphing, you could simply transpose the voice upward a bit, as most (non hearing impaired) people find female voices easier to understand.

This can be very time consuming and PCs have only made that part of it worse. To get the best result, you have to play with it a lot.


I have a feeling that you're preparing it for a speech or something. Do people need to hear every word or could most of it be drowned in music or "pleasent sounds" while following a transcript on a screen or paper (most people will "hear" it even if they can't, as long as they have a transcript - the human mind is good at such tricks).

If David Copperfield could score Claudia, then you should be able to smoke and mirror your audience 

And yea, I had no idea during production that there was a background static noise. A proper recording wasn't possible given the situation. Otherwise I'd just redo it.

Zoom mikes were born for this, and I have considered a simple (but stable) bug for this kind of takes - just place it under/right in front of the stage (with blue tack or chewing gum) and have a small good quality receiver at your recording device.

Well, and a time machine to fix the present problem of course.

Best of luck.

[Admin]

I saw this robot at a museum . . . it talked . . . all I had on me was my trusty digital camera. I just want to put up ~30 seconds of it on youtube. In this situation, altering the voice, adding background music, or refilming it isn't available options . . .

But yea, if it was any other situation, I'd do it!

When I filmed my Axon II sonar demo example, it took me like 10 retries until I realized that the sonar somehow negatively interacts with my camera speaker. For the life of me I couldn't figure out why sound wasn't working until like the 6th attempt. In the end, I ended up dubbing over the video in post production.

But yea, isn't an option here, I need it true to what it actually sounds like. Soeren I tried my Sony Vegas Studio software to remove bands (it has sliders), and it appears the noise is in the voice bands as you said . . . so probably can't remove the noise without a much more advanced noise remover algorithm.