removing "hum" from an audio recording

I tuned in late, but for audio processing free of charge, try GoldWave.

I have the version referenced in that web page if you want it. Its

8.24 MB. E-mail me if you want it and I'll stick it up on one of my websites for you.

--
Service to my country? Been there, Done that, and I\'ve got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

Try moving the cutoff frequency higher. At some point it will make the voices sound funny, but you can definitely go higher than 100 Hz.

--Mac

The pass band for telephone audio is 300 Hz to about 3 kHz. As I understand it, this wasn't (originally) because of any difficulty in having a wider pass band, but because that's where most of the information in speech is, and removing the other frequencies helps intelligibility by removing distracting background noises.

So if the purpose of recording the meeting was just to hear what people said (no music or sound effects :-) ) then you might try filtering out everything below 300 Hz and above 3-4 kHz anyway, and see if that makes the recording more intelligible, even aside from the hum problem.

--
   Wim Lewis , Seattle, WA, USA. PGP keyID 27F772C1

If you have access to an adjustable notch filter you might be able to get rid of the hum if it is at one frequency.

How do you know that it is not your tape reader that is producing the hum? Have you tryed a second machine?

al

Prior to moving the machine, everything sounded crystal clear.

Then I moved it, then we started unexpectedly.

Then on playback, hum...

Oh well.

I tried the "Noise Removal" in Audacity; hum was gone (yay!), but now the voices sound a bit like aliens.

Management told me not to "waste any more time on cleaning it up". Go figure.

Thanks all who replied...

Michael

I'll bet you sampled the whole recording and subtracted that out, leaving behind, at any moment, only the difference between the instantaneous spectrum and the average of the whole thing.

Instead, sample a portion that contains only noise (not voice) and use that as your noise sample. I don't know how to do that in Audacity, but it should be fairly straightforward. For GoldWave instructions see:

You should get beautifully clear results.

Er... I took a small sample of nothing but noise, told Audacity to recognize this as noise, then highlighted the whole 30-minute clip and said, Go Ahead and Remove the Noise.

The problem is, once someone starts talking, the amplitude of the hum decreases. So, I'm guessing the algorithm is trying to subtract hum from audio patches which have no hum - resulting in an alien-like effect.

It's ok, it's quite possible we won't even use the recording, since we have the questions in writing... (and have to provide the answers in writing as well...)

Michael

Ah.

Could well be. You might try telling it to subtract only 50% of the noise, or something like that.

Look at the slider, don't go over the yellow range in the middle. You didn't follow exactly my instructions there. You can preview the result and put much less effect into the final. You can also modify the fft settings. uncheck "live update" and set the fft size to 8192. Click on help and read about the different parameters and how they affect the outcome. Experiment a bit before you complain here.

--
ciao Ban
Apricale, Italy

It sounds like it has AVC on the record part. Almost like the recorder needs a "squelch" control. ;-)

But a 300Hz low-pass will get rid of a lot of crap - people will sound a little nasal, but not alien. Oh, and putting this filter on the input of the recorder will keep the crap off the tape in the first place.

Good luck! Rich

Right. My experience, however, having done some speech research, is that it helps if you go up to about 5 kHz instead of just 3 kHz. A strict cutoff at

3 kHz will sound just a bit muffled.

It should.

Well, good grief! You could find a studio, and read them again! ;-)

Cheers! Rich

Read the help file

The depth of noise reduction is to great leaving electronic artifacts. Reduce the amount of noise reduction until it is as best as possible and without the artifacts. Test on a small section first to save time.

You could also try the notch filter. Look at the spectrum of noise and try and determine if there is a single frequency that relates to your hum.

Peter

I did the high pass filter from 300 hz, and the hum is mostly gone now, thanks.

However, I still have these strange "peaks" of noise superimposed on my signal. There would be a peak maybe 1 cm tall, then about 2 inches later, a peak of opposite sign will show up. Seems immune to the low-pass filter. It's not a regular sine wave... anyone familiar with what these are?

We're not really supposed to be installing software on our work computers, no notch filters. I used Audacity because it bypasses the Windows installer.

now

leaving

use that

it

try

Audacity is a good tool, gets better as you learn it more. First observation the original signal has already been compressed (audio / instrument sense first), this interferes with the relationship between the noise and the desired signal. Noise reduction / recognition / removal works best on un-compressed signals. Try various Dolby as well as DBx uncompression models. If done correctly the noise properties should match.

--
JosephKK
Gegen dummheit kampfen die Gotter Selbst, vergebens.  
--Schiller