Reducing Audio Noise

You may have encountered the problem that your video contains unwanted back ground noise. This article will discuss the basics of audio noise reduction and two ways of getting rid of it.

Types of noise

System intrinsic noise

In electronics, noise is defined as any unwanted electrical signal, unrelated to the original signal. It appears in every electronic circuit. Recorded sound whether stored as analog signal or as digital data all has gone through electronic circuitry that converted the sound to electrical signals. This means that recorded sound will also always contain a certain amount of noise. Also tape is not perfect: it will also always contain some level of noise.

External noise

Besides noise as defined above, in audio you can also find noise that is part of the original signal. Examples of this type of noise is when you record in a room with an loud air conditioning or inside a driving car or flying aircraft. Also the noise made by the electrical motors in your camera that drives your tape or the zoom mechanism may be audible in silent passages. But you can also extend the definition of unwanted sound to situations where, for example there is a passing car. Sometimes you may consider it as part of the natural sound, sometimes you don't.

Basics of noise reduction

For system intrinsic noise all kind of systems have been developed to reduce it. Very well know techniques for noise reduction are those that are developed by Dolby Laboratories. They rely on the fact that the noise is very uniform and predictable.

But what if the noise is not predictable? In this case the general principle is to try to analyze the signal over a period of time and recognize the pattern that makes up the noise and filter it out. The most simple form of filtering out is by finding out which frequencies make up the noise and then simply remove or reduce the volume of these frequencies. This however has it's limitations because it will also influence wanted sound that also have these frequencies.

A more sophisticated way of filtering out is subtracting a signal that is exactly equal to the signal that is determined to be the unwanted noise. This is a very effective way and you can reach an amzing good result with it. This method works best if you have a piece of footage where you only have the noise, so that you can sample it and substract. Let's have a look at two packages that you can use in order to demonstrate how it works.

Adobe Audition

This very powerfull audio editing program is the further development of another program Cool Edit, developed in the past by a company called Syntrillium Software. If you're a very serious in audio editing, this is a very good choice. Another advantage is that it integrates seamlessly in Adobe Premiere.

But let's have a look how you can use it for noise reduction. We have prepared a short segment with some human sounds with a strong background noise (originating frorm a defective camcorder). Click here to hear the segment. Below there is a screenshot of the segment opened in Audition. Listening to it and looking at the picture you can easily identify what sound is at what position.

To learn the program what noise is we have to select a segment where there is only noise and no wanted sound. In this case in the beginning there is such a segment. Selecting this segment by dragging the mouse over it and open the Noise Reduction filter on the Effect panel will open up the next window.

In this window there is a simple button "Get Profile from selection". What simply happens is that the program will "learn" that this is noise and store it in a profile. If you have done that you can simply Close the Window. (don't push OK, because that will directly apply the profile to the selection.

Arriving back in the main window you now have to select the whole segment where you want to reduce the noise. Now you have to open the Noise Reduction Filter again which will bring you back to the sreen on the left again. But this time the profile that was just created is automatically loaded. You can push the Review button to listen to the result. There are some settings that you can play around with to influence the result, of which the most important one is the slider Noise Reduction Level. After pressing OK you have applied the profile to the whole selection.

On the right you can see the how the file looks like after the noise has been reduced. It is clearly visible that the stretches where the original file contained little sound have been reduced considerably, while the spots where sounds are audible have hardly changed. You can listen to the resulting sound here.

Soundsoap

This program was designed by the US based company Bias (www.bias-inc.com). They supply a range of audio editing products products. The Pro version can be used as a plug-in in most video editing programs as long as they allow application of DirectX filters and offer more tweaking than the normal version.

The normal version can also be used as plug-in, but also as a stand-alone program. Although this has the disadvantage that you will first have to extract the audio from your video file, reduce the noise and put it back into the video project, we'll have a look at the stand-alone version. It demonstrates to what level of the principle of noise reduction can be automated. Below you see a screenshot of the extremely simple user interface.

Through File -> Open you can open the sound file. By pushing the play button, you can listen to it. I used the same file to demonstrate what the program is capable of. Once you have opened the sound file, the only thing that you have to do is to activate the Learn Noise button and press the play button. If so desired you can define an In-point and an Out-point if you only want to treat a part of the file. If you now press the play button again you will here the result. With the Noise Tuner and Noise Reduction knobs you can tweak the noise reduction. Interesting feature is that it also has a Preserve Voice button. Human voices tend to have certain characteristics and Soundsoap has an algorithm that reduces the impact of noise reduction on the color of human voices. Furter more it has a rumble amd 50 and 60Hz hum removal.

You can listen to the result sound file here.