1

My colleague and I attended a conference and made a number of interviews in a crowded hall full of people.

We are completely new to video recording: it was just Panasonic HC-V777 camcorder (~300$) and nothing else. We got great content, great picture and good enough voice of speakers but also a terrible background noise - sometimes the conference room was full of people and that makes it hard now to listen and understand the people on some of our videos.

We are mainly Mac OS X users so we tried iMovie and its "Reduce background noise" with different equalizer presets it suggests. This has some effect on our videos but still the ratio speaker vs crowd is not satisfactory.

We are wondering if there is anything more powerful for this kind of noise reduction - specifically reduction of background noise produced by crowd - tens or hundreds of people talking to each other in a big hall.

  • Also, consider that what you're trying to reduce or eliminate isn't "noise" in the technical sense -- it's audio in the same range and at almost the same levels as your desired sound. It's only noise in the same sense that weeds are plants you don't happen to want around. Most strategies for noise reduction start with the premise that there's some critical difference between the wanted and unwanted signals, then identify and exploit that difference. What would that be in this case? – Jim Mack Apr 11 '17 at 2:23
3

There are automated algorithms like those you tried in iMovie, the ones in Audacity might be worth a try but if you want better results you should individually edit phrases.

Software like izotope RX can help in the restoration process.

  • 1
    +1 for Izotope, I've never found anything better... but you need deep pockets & it's not always perfect. Test a demo if they have one, before committing. – Tetsujin Apr 11 '17 at 17:35
  • 2
    I'll +1 Audacity because it has a way to grab a bit of the background and use that as the filter. It's free and worth a shot, anyway. youtube.com/… is a good start (first one explains the steps well) – Hans Apr 12 '17 at 15:01
1

If you have a stereo recording, your speaker has a certain position with regard to the microphones. Now your stereo base for a camcorder is quite small, nevertheless you can average short-time cross correlations to get a cue for the direction of the initial, correlated wavefront (uncorrelated stuff averages out), then create a Wiener filter based on the ratios of correlated-as-expected and loud-as-expected and other sound and implement it with overlapping windowed FFT filtering.

This can be surprisingly effective in quieting background noise at the cost of introducing warbling and blubbering "musical noise": basically the comprehensibility of the main speaker improves while the overall sound becomes distracting. You can tune that sort of algorithm to get good tradeoffs.

When the original recording is mono, you can basically can only work with signal strengths and use stuff like frequency-selective noise-gating.

That's decidedly less effective and it doesn't help that most of the background noise will be speech with characteristics similar to that of the speaker.

-1

There are some tools you can use from Audionamix to help reduce unwanted background noise. I have used a plugin called VVC (Vocal Volume Control) and IDC (Instant Dialogue Cleaner) to help reduce background noise. IDC works as a realtime plugin for a DAW and after testing it was impressed with the results it gave. You can watch a demo video here

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.