0

Due to excessive use of noise reduction, I can hear audio clipping (in headphones).

Is there a way to repair the audio after the noise reduction has been applied, so that it doesn't clip? Here's a recording.

  • Can you upload an example? I don't know what an "audio chip off" is, but I bet I'll recognize it if I hear it. – Todd Wilcox May 9 '17 at 13:10
  • @ToddWilcox, I had a link to the sample above in "repair the audio" phrase, but it seems less visible. I don't know if to call it chip off, but they just sound like one.(filebin.net/i4w8gbcb4eklxq5g) – lind May 9 '17 at 15:06
  • "chip off" = "clipping"? – Yorik May 9 '17 at 17:05
  • @Yorik, I think so. :-) – lind May 10 '17 at 1:22
  • I listened to your sample, and I didn't detect any clipping. What I did hear was very aggressive noise cancellation. The issue with this, and clipping, is that repairing it is difficult if not impossible. With noise reduction, you've used a tool that has analysed the constant background noise and audio artifacts in your recording and it has erased them. This process also damages the audio you wish to preserve; in this case a human voice speaking. Therefore you are trying to put back something that has been removed. I would recommend using less aggressive noise cancellation. – ABragg May 10 '17 at 9:46
0

Firstly, I agree with ABragg that it's not clipped. I downloaded and zoomed in graphically and didn't see clipped waveforms.

But, to answer your question (which may attract other readers):

Clipping is by definition the loss of information. It looks like someone has chopped off the waveforms at a certain level.

In general, you can't get back what was lost. In theory, however, you can reconstitute it plausibly (meaning that it will sound passable) by guessing at the lost info. This can be from a detailed analysis of past waves when the volume was lower, or simply by a bit of rounding guesswork based on local sampling data.

This sort of thing is easier to do when you have a variety of dynamics with a similar waveform -- for example, a piano has fairly similar timbre at levels beneath what is clipped, as opposed to an organ with no reverb.

This tool claims that it can help with clipping, and here's a video demo.

I might as well also throw in a link to the FBI's investigation into this, in 2010, for any who are interested.

I would not be too surprised to learn, now or in the near future, that someone has come up with a means of applying a pretty effective AI algorithm which learns a general range of waveforms and uses this to fill in clipped waveforms, even from a variety of sources such as within an orchestra. This sort of thing has been done with some success in graphics.

  • would the sample above be considered a good quality audio? – lind May 10 '17 at 15:05
  • what are the ways to enhance the sample as such that I have posted above digitally? – lind May 10 '17 at 15:05
  • Do you mean the recording you provided? It sounds a bit distorted, but I don't know the cause. It could be compression (by which I mean digital compression, as with an MP3, not audio compression, which is a way of making the volume peaks and low points closer in volume), but it could also be feedback, such as if you speak near a speaker and the sound feeds back from the speaker into the mic. Both of these can create one or more resonant frequencies, and I hear some of that in your sample. Maybe someone else here will recognize the exact nature of the distortion better than I can. – Epanoui May 10 '17 at 16:58
  • To answer your second question: The first thing which I notice is the unwanted frequencies, which you can reduce with selective equalization. For this you need to apply an equalizer (also called EQ) and set a narrow peak, then move it around to find the frequency you'll want to reduce -- when the unwanted sound is much louder, then you've found it, and you will then reverse the EQ's peak to negative to deemphasize that frequency. It should be narrow to affect little more than the unwanted frequency. You can do this with multiple frequencies. Samplitude is a program with this ability. – Epanoui May 10 '17 at 17:01
  • Further, there's some noisiness. You should try to remove the frequencies first, then go for the general noise. Many tools have noise reduction. It just occurred to me that Audacity is an easily accessible audio editing program which will be quicker to install and should be effective. You may need to experiment and stay mindful of overall sound quality -- it's possible to take all of something out, but in doing so to become so focused on the adjustment that you don't notice the overall change in sound quality. Sometimes the best result is a compromise which still sounds natural. – Epanoui May 10 '17 at 17:04

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.