NGWave Audio/Sound/MP3 Editor: Noise Reduction Part I

NGWave's Noise Reduction can be quite useful in reducing background noise, tape hiss, and other unwanted noises. This tutorial walks you through a session of cleaning up an older analog recording, removing as much hiss as possible, but without reducing treble response, or introducing artifacts into the sound.

First, a bit about how the noise reduction works.

FFT Simplified

Like many functions in NGWave, the Noise Reduction utilizes an FFT, or Fast Fourier Transform. Without getting too far into the technical details, an FFT simply transforms the audio into the frequency domain.

Depending on your FFT settings, when a portion of audio is transformed into the Frequency Domain, each individual frequency is represented. This is very different from the more standard representation of samples.

We will assume the default FFT settings are being used (2048 Window Size, 8x Overlap) for the duration of this article.

Transformation

First, NGWave converts your audio into a set of 1024 distinct frequencies. Each band has a width of about 21.5 Hz. Internally, NGWave knows what frequencies fall within each band, and what the phaze and level is for each frequency, down to a 21.5 Hz resolution. Larger window sizes result in more frequencies; for example, a window of 4096 samples gives a resolution of about 11 Hz.

This FFT representation of the audio is used in NGWave's FFT Filtering. Quite simply, if you want to simply cut off any frequencies above a certain value, it zeros those values, and then converts the audio back into the Time Domain, which is the typical Sample representation. The result is audio with the higher frequencies removed, or filtered.

Noise Reduction

To apply Noise Reduction, NGWave applies a Noise Gate to each frequency band. Tape hiss, for example, is generally high-frequency noise, at a very low level. By zeroing out any frequencies below a certain threshold -- and, more importantly, allowing frequencies above that threshold to come through at full volume -- we can reduce background noise efficiently.

Note that this is very different from a typical broadband noise gate, that simply turns the audio on or off depending on its current level. Noise Reduction applies a noise gate to each individual frequency, and is much less noticable if used properly.