Announcement

Collapse
No announcement yet.

Auto Spectrum CNF vs. Forensic AFDF

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Auto Spectrum CNF vs. Forensic AFDF

    Recently, we received a question from a Live-Forensics user regarding the differences between the Auto Spectrum CNF compared to the Forensic AFDF (Adaptive Frequency Domain Filter). I thought that the answer to this question may be of general interest.

    Both the Auto Spectrum CNF and the Forensic AFDF are adaptive filters found in the Continuous Noise Filter. Both filters take their own cascades of noise fingerprints on the fly and adjust those fingerprints appropriately to the noise measured throughout the file. But, they are two different algorithms and have different sonic performance characteristics. The Auto Spectrum CNF is optimized for musical material having higher values of S/N (signal to noise ratios) and wide bandwidths while the Forensic AFDF is optimized for poor quality speech in which the S/N is very low or near unity and the frequency response of the material is more limited.

    The question went on as to whether or not the Forensic AFDF could be used on musical material. The short answer is yes, but it is not generally the optimal filter to use for that. It has a faster transient response than the Auto Spectrum CNF but produces higher levels of artifacting. However, we have heard of customers using the Forensics AFDF for some musical audio work (cylinders and early acoustical 78's), but the key is to keep the attenuation control to a pretty low setting to avoid artifacting. I will go on to say that the Auto Spectrum CNF does not make a very good Forensics filter when the S/N and signal intelligibility are very poor. However, if you do not have the Forensics version of the software and need an adaptive filter for some Forensics type of audio restoration, it is worth trying out.
    Last edited by Craig Maier; 06-10-2006, 08:39 PM.
    "Who put orange juice in my orange juice?" - - - William Claude Dukenfield
Working...
X