The BNR Spectral Graph displays the noise signature and the editable noise Contour Line. The Spectral Graph’s horizontal axis shows frequency, which is displayed in Hertz, from 0 Hz to one-half the current audio file’s sample rate. The Spectral Graph’s vertical axis shows amplitude, which is displayed in dB, from 0 dB to –144 dB (below full-scale output of the audio).
The jagged line is a graph of noise. This is called a noise signature. It is created when you use the Learn button in the Broadband Noise Reduction window. Once you have the noise signature of an audio file, you will be able to begin removing the noise by generating and editing a threshold or Contour Line (covered next) between the noise and the desired audio signal.
The line with a series of square breakpoints is called the noise contour line. The Contour Line is an editable envelope which represents the division between the noise and the non-noise signal in the current audio file. The Contour Line is created by clicking the Fit or AutoFit button in the Broadband Noise Reduction window after you have learned a section of noise. By moving this envelope up or down, or by moving the individual breakpoints, you can modify which signals are removed and which remain.
The noise modeling process treats audio below the line as mostly noise, and audio above the line as mostly signal. Therefore, the higher you move the Contour Line upwards, the more audio is removed. To maximize noise reduction and minimize signal loss, the Contour Line should be above any noise components, but below any signal components.
To fine-tune the broadband noise reduction, move breakpoints at different locations along this line to find out which segments remove the noise most efficiently. Editing the Contour Line to follow the noise signature as closely as possible will also help maximize noise reduction and minimize signal loss. See Editing the Contour Line.