Noise normalization

From Hydrogenaudio Knowledgebase
Revision as of 10:54, 23 November 2017 by (Talk)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Noise normalization is a special type of coding technique (somewhat comparable to PNS used in AAC, but not the same) used by Ogg Vorbis psychoacoustics model to compensate for loss of energy in certain frequency bands due to quantization (usually because of a low SNR). The algorithm uses an apsort function to determine where the energy lost in each frequency band is the greatest and compensates the band with the greatest energy lost by reviving the residue (noise error) (see psy.c in source, _vp_noise_normalize_sort function, line 1050). This is main cause of most metallic artifacts in MP3. It is one factor that gives Vorbis its noise-like sound characteristics, when the bitrate is too low to encode without susceptible artifacts. In the reference encoder it's used on all quality values less than 4.