User`s manual

114
4.5.3: Adaptive Spectral Inverse Filter (ASIF)
Application:
The Adaptive Spectral Inverse Filter (ASIF) is an equalization filter that automatically readjusts
the spectrum to match an expected spectral shape. It is especially useful when the target voice
has been exposed to spectral coloration (i.e. muffling, hollowness, or tinniness), but it can also
be used to remove bandlimited noises. This filter is much like the Spectral Inverse Filter (SIF,
See Section 4.4.12: ), except it continually updates the spectral solution, whereas the SIF only
updates the solution when it is “built”.
The ASIF maintains an average of the signal’s spectrum and uses this information to implement
a high-resolution digital filter for correcting long-term spectral irregularities. The goal of the
filter is to reshape the overall spectral envelope of the audio, not to respond to transient noises
and characteristics.
Several user controls are available for refinement of ASIF operation. The user can specify the
expected spectrum so that the output audio is reshaped to a flat, pink, voice-like, or custom
curve. An adapt rate setting controls the update rate for the spectral average, which in turn
determines how quickly the filter responds to changes in the input audio. Upper and lower limit
controls allow the user to specify the range over which equalization is applied, and a mode
setting controls whether frequencies outside the equalization range are attenuated or left
unaffected. The amount of spectral correction is adjustable using the Filter Amount control.
The user can enable the auto-gain functionality to ensure that the output audio level is
maintained at approximately the same as the input audio level. If the user disables the auto-
gain, an output gain slider is available to manually boost the level of the output signal.
As an aid to visualizing the filter operation, the user can choose to view the input and output
audio traces or the filter coefficient trace.