Removing humming sound during talk

[AjnabiZ]

Hello

I have a clip where there is a sort of humming sound as the recording was done in a Hall.

How do i remove the hum during the voice/talk audio.

[davexnet]

You use an audio editor, such as Audacity, and play around with the noise reduction tools.

[Cornucopia]

Alot depends on the kind of noise the hum is.

Most likely, what you're looking for is a FILTER.
If the hum is AC 50/60Hz line hum, a simple Highpass filter with a cutoff frequency of ~100Hz will do a good job.
If it's a different kind of hum it could be more complicated. If there are harmonics of the hum's fundamental frequency, add a comb filter. If the hum's fundamental is higher in frequency, use a notch (band-cut) filter instead of a highpass.

Though there is always a certain amount of "playing around" or adjustment needing to be done in these situations, it's best if you approach this using scientific/deterministic methods. For example, to find the closest approximation of the hum's fundamental - use a reverse filter (set to boost). You'll notice immediately how the hum blooms/drones/pops when you get to that resonant frequency. Then reverse the filter back again.

Scott

[budwzr]

This is a good example of why you want to get it right the first time. Since it was done in a hall, Scott is saying use the "Hall" effect and reverse it. Is that right Scott?

That's certainly an interesting approach. Hmmm. Thanks for posting that. It's got me thinking.

[Cornucopia]

Theoretically, if the HUM were constant and predictable, you could fairly easily remove it. Kind of like an equation:

Signal + Noise = SignalNoise, SignalNoise - Noise = Signal

The problem, of course, is that it's NEVER that constant or predictable. HUM wavers (both cyclically and randomly) in frequency, loudness and phase, plus there would be nearly infinite room/hall echoes of that HUM arriving at different angles (and phases, and delays) for it to be reasonablly removed. Then it's more like this:

Signal + Noise = SignalNoise, SignalNoise - Noise2 = SignalNoise(inverted)Noise2 --- this is WORSE!

No, what I was suggesting was that if ALL you want is the human voice, which is a fairly narrowband signal (~85Hz--5000Hz for the great majority of the fundamentals and harmonics), you can remove lower pitched sounds like deep bass (and 50/60Hz hum) with a highpass filter (aka Low Cut), and if there's HISS, you can remove some/much of that with a lowpass filter (aka High cut), leaving just the middle range.

Often though, HISS is wider ranging in pitch than that, so you have to try multiple techniques. HUM from REALLY BAD AC devices starts at 50/60 Hz, but has loud harmonics (2x=100/120, 3x=150/180, 4x=200/240...) and so a COMB filter (which can basically be achieved with a VERY FAST delay) will deeply remove many of those harmonics exactly with very little loss in the frequencies AROUND those pitches.

Of course, it's always best to have your recording processes such that there is no NEED to fix something in post - hopefully, the signal would be clean enough to begin with. Remember, in addition to correcting what's already been done, take the time to analyze the setup to see how to avoid hum in the future (including bringing the mike CLOSER to the target, making the mike more DIRECTIONAL to the target and less sensitive to surrounding sounds, use a mike whose sensitivity spectrum more closely matches the vocal range and little else, isolating/shielding the path from mike to recorder, applying materials to dampen the room echo of the hum, and of course minimizing/stopping the hum source itself). This analysis can also sometimes be used to "model" the way the hum is generated such that you can cancel/reverse/undo the hum sound(s).

Another REALLY COMPLICATED thing to try is convolution:

Take a recording of room silence (without the hum - make sure it isn't there).
Take another recording of that same room (with the same recording parameters & levels), but this time include the hum.
Get impulse response graphs from both recordings and see what the impulse graph would be for the difference between the two.
Apply that Difference graph as a convolution operand to the signal+noise.

Basically works as an extremely well fitting and finely tuned set of filters.

Scott

[AjnabiZ]

Would i need to purchase Noise Reduction Plugin for Sony Audio Forge 10 ? or does it have it built in ?

[AjnabiZ]

Theoretically, if the HUM were constant and predictable, you could fairly easily remove it. Kind of like an equation:

Signal + Noise = SignalNoise, SignalNoise - Noise = Signal

The problem, of course, is that it's NEVER that constant or predictable. HUM wavers (both cyclically and randomly) in frequency, loudness and phase, plus there would be nearly infinite room/hall echoes of that HUM arriving at different angles (and phases, and delays) for it to be reasonablly removed. Then it's more like this:

Signal + Noise = SignalNoise, SignalNoise - Noise2 = SignalNoise(inverted)Noise2 --- this is WORSE!

No, what I was suggesting was that if ALL you want is the human voice, which is a fairly narrowband signal (~85Hz--5000Hz for the great majority of the fundamentals and harmonics), you can remove lower pitched sounds like deep bass (and 50/60Hz hum) with a highpass filter (aka Low Cut), and if there's HISS, you can remove some/much of that with a lowpass filter (aka High cut), leaving just the middle range.

Often though, HISS is wider ranging in pitch than that, so you have to try multiple techniques. HUM from REALLY BAD AC devices starts at 50/60 Hz, but has loud harmonics (2x=100/120, 3x=150/180, 4x=200/240...) and so a COMB filter (which can basically be achieved with a VERY FAST delay) will deeply remove many of those harmonics exactly with very little loss in the frequencies AROUND those pitches.

Of course, it's always best to have your recording processes such that there is no NEED to fix something in post - hopefully, the signal would be clean enough to begin with. Remember, in addition to correcting what's already been done, take the time to analyze the setup to see how to avoid hum in the future (including bringing the mike CLOSER to the target, making the mike more DIRECTIONAL to the target and less sensitive to surrounding sounds, use a mike whose sensitivity spectrum more closely matches the vocal range and little else, isolating/shielding the path from mike to recorder, applying materials to dampen the room echo of the hum, and of course minimizing/stopping the hum source itself). This analysis can also sometimes be used to "model" the way the hum is generated such that you can cancel/reverse/undo the hum sound(s).

Another REALLY COMPLICATED thing to try is convolution:

Take a recording of room silence (without the hum - make sure it isn't there).
Take another recording of that same room (with the same recording parameters & levels), but this time include the hum.
Get impulse response graphs from both recordings and see what the impulse graph would be for the difference between the two.
Apply that Difference graph as a convolution operand to the signal+noise.

Basically works as an extremely well fitting and finely tuned set of filters.

Scott

I did not get what you said, im a beginner in audio recording and editing.

[budwzr]

I'm familiar with convolves in video pixels, never thought about it for sound. Whew, that's getting out there into the esoteric.

[Cornucopia]

If you're already considering SoundForge (which is what I think you meant), then you should be able to do most of what I said with the provided plugins that come with SF. There are other 3rd part plugins that might do more complicated and exacting things, but SF is quite capable for most Audio needs (including Hum Removal).

If you don't know how to use it, read the manual and internet/user tutorials, and after that if you're still unsure, come back and ask here - you'll then be able to explain the situation in more detail.

Scott

[AjnabiZ]

If you're already considering SoundForge (which is what I think you meant), then you should be able to do most of what I said with the provided plugins that come with SF. There are other 3rd part plugins that might do more complicated and exacting things, but SF is quite capable for most Audio needs (including Hum Removal).

If you don't know how to use it, read the manual and internet/user tutorials, and after that if you're still unsure, come back and ask here - you'll then be able to explain the situation in more detail.

Scott

Thanks Scott


yes, i plan on using Sony Sound forge that is why i opened this thread

https://forum.videohelp.com/threads/340090-Is-there-a-built-in-Audio-editor-in-Vegas-Mo...tudio-Platinum

[Cornucopia]

See the other thread for an answer to that last bit...

Scott

[AjnabiZ]

See the other thread for an answer to that last bit...

Scott

Which thread ?

[davexnet]

If you get the full Sound Forge, ($374) Sony noise reduction is included.
However, if you get the "Studio" version ($64) it is not.

The way you use the NR, with the noise print, and then adjust the level and/or threshold,
is pretty much what Cornucopia described earlier:
Signal + Noise = SignalNoise, SignalNoise - Noise = Signal

I have SF 8 with Sony NR, and I can say that it works quite well with many kinds of problem audio
It has a preview mode where you can switch it to listen to just the audio that is being removed.
If you hear noise and some signal (voices, for example) you can back it off until the only the noise is heard
and that's a good place to start.

[lordsmurf]

So many words in this thread.

Upload a sample clip. I'll quickly play with it, see what's possible.
(PM me when that's done.)

[AjnabiZ]

So many words in this thread.

Upload a sample clip. I'll quickly play with it, see what's possible.
(PM me when that's done.)

Thanks Lordsmurf, just sent you the PM with the original audio file with noise.