How to Reduce Background Noise in a Voice Recording

How can you improve an audio recording that contains excessive background noise? Many audio editing software programs include noise reduction tools or plug-ins. The free audio editor Audacity supplies an effect called Noise Removal. When and how should you use this tool to clean up a spoken-voice recording?

What does Noise Removal remove?

To effectively use Audacity's Noise Removal effect, you must understand what it will remove. Like any noise reduction tool, Noise Removal will not remove temporary or changing sounds such as those caused by traffic, pets, or voices in the background. These sounds are best treated as mistakes on the recording. See Amateur Voice Recording: Audio Editing by Selecting and Deleting for more information.

Noise Removal will reduce consistent, pervasive background noise. If you hear a hiss, hum, or buzz that doesn't change throughout a recording, then you can improve the sound quality with Noise Removal.

How do I use Noise Removal?

Noise Removal is somewhat different from most Audacity effects and plug-ins. It can, therefore, confuse a new user. Starting with version 1.3.3, the effect is more complicated than in earlier versions, although it provides much better noise reduction. Let's go over the controls, from top to bottom.

Get Noise Profile

Before Audacity can remove noise, it must know what the noise is. Select a part of the recording that would be pure silence if not for the background noise. The selection should not contain any mouse clicks, breaths, voices, or other sounds. The longer the selection is, the better.

Choose "Noise Removal" from the "Effect" menu. Click "Get Noise Profile." You have now taught Audacity what the background noise in your voice recording sounds like. Select the entire recording, then bring up the Noise Removal dialog again.

Noise reduction (dB)

Noise Removal in older Audacity versions used a single slider that ranged from "Less" to "More." The newer "Noise reduction (dB)" control serves the same purpose as the older slider. This determines the amount to reduce the noise by.

If this setting is too low, you will keep unwanted hiss or hum in the recording. If it is too high, it will introduce artifacts - new sounds that were not initially in the recording. These artifacts usually create a warbling or burbling sound in the background.

Start with this control very low. Preview the processed recording. If the background noise level is still too high and no artifacts are present, slightly increase the noise reduction amount. If you hear any artifacts, decrease the amount. When you reach the highest level of noise reduction that does not create artifacts, move to the next slider.

Frequency smoothing (Hz)

Frequency smoothing prevents noise reduction artifacts by blurring the noise profile. It also causes Noise Removal to reduce some parts of the voice recording that are not background noise.

If the frequency smoothing control is too low, even low amounts of noise reduction will create artifacts. If it is too high, the voice recording will sound muffled or distorted.

Start with frequency smoothing around 200-300, and preview the processed recording. If you hear any artifacts, slightly increase this control. If the recorded voice sounds distorted, decrease the control.

Attack/decay time (secs)

In principle, the attack/decay time prevents artifacts at the expense of slow transitions between voice and silence. In practice, for a simple spoken-voice recording, you can leave this control at its lowest setting.

When am I finished?

Increasing frequency smoothing lets you increase noise reduction without artifacts, but increased noise reduction allows less frequency smoothing without distortion. Because the controls affect each other, you may need to adjust each one several times before reaching the best combination of values. When you find the highest amount of noise reduction possible with absolutely no audible artifacts or distortion, you are done adjusting the Noise Removal effect.

Audacity's Noise Removal is one of several ways to reduce background noise in a voice recording. When used carelessly, this effect can ruin a recording with artifacts and distortion. If you apply it appropriately, however, you can enhance a the quality of a voice recording.