Noise Removal

Noise removal is accessed via the Processings/Noise Removal menu.

Select an item in the pop up menu.

• Energy Substraction : the level of noise to be removed is estimated in terms of energy.

• Amplitude Substraction : the level of noise to be removed is estimated in terms of amplitude.

With the second method, more noise tends to be removed.

Two parameters are at stake : noise reduction, and overcompensation. These paremeters define the application of the fingerprints with time varying thresholds.

• Noise Reduction : maximum amplitudes in dB of each frequencies observed in the noise zone

• Overcompensation : variations of these amplitudes. This is a factor, that allows to increase or descrease the threshold in the case important noise variations. The higher the factor, the more peaks are substracted.

• Suppress Sinusoïdal Noise : sinusoids whose level is lower than that of the detected noise level are removed prior to applying spectral substraction.

• Suppression Attack Time : a value in ms, which specifies the time constant in ms for the attack of the spectral subtraction filter. This allows to preserve the reverberation of previous sounds for instance.

Check the Normalization option in the dialogue window. This option is only applied when processing the sound. Normalization can't be heard at preview.

• To get a preview of the filtered sound, use the corresponding button.

• To activate the function, click on the preview button once. The function will be activated, so that you can hear the sound without and with the noise removal alternativelay.

• To listen to a loop of the sound, use the loop button.

If a single noise zone is located in the file, its values will be applied to the whole file, or to the selected part of the file.

If several noise zones have been created or imported to the file, they will be interpolated dynamically. The interpolation then results into a dynamic fingerprint : the noise removal level will change with the temporal location in the signal.