[vorbis-dev] bark_noise

[Segher Boessenkool]

Hello dev people,

Can someone please explain me why bark_noise works so well?

I can't find any references to theory. I think Monty just
made it up (in the Netherlands we say: he sucked it out
of his thumb :-) ).

Especially why the 1/|frequency distance| thingie is a
good spreading function. I tried several different ones,
but the original is still the best overall. Some of the
alternatives created a (artificially) clearer sound
(with some twinkling artifacts), and some made a (much!)
better noise rendition, at the cost of tonal quality.

But the original wins, in my (listening) tests.

I hate this, as bark_noise is one of the routines that
needs optimizing. If I understand _why_ it works, maybe
I can optimize it to something different, that performs
just as good in sound quality, and better in execution
time.

A second question: is there a standard way (like, executing
some obscure program in the vq/ directory) to create
_cascading_ codebooks? (I only need additive codebooks,
not multiplicative ones). And is it possible to have the
second-level and third-level etc. codebooks a different
grouping-length than the first-level ones?

Oh, and I want a _perfect_ 40kbps mode before the end of
the year ;-)

Cheers,

Segher

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.