On Wed, Feb 29, 2012 at 5:19 PM, Sampo Syreeni de...@iki.fi wrote:
Personally what I find a bit worrisome is that this sort of optimization
retains the blackbox leanings of machine learning as a general discipline.
None of the ambisonic specific, closed form optimization literature, or the
derived specifics of the base optimization problem, are being utilized.
Instead the two (sometimes simultaneous, sometimes even not that) Gerzonian
equations are being fed into one or another optimization framework, with no
regard to what happens then, and without feeding in all of the age-old
mathematical-physical knowhow of how those systems of equations behave. Like
for instance psychoacoustical sensitivity estimates from the BBC era.
Indeed. But in 100 lines of fairly simple code or so — you have
something that produces usable matrixes.
I'm also a little bit of a skeptic towards the
stuff. At least as far as the math I know and love suggests I should be.
It's not magic, for sure— but the problems we solve here are small
(few dimensions), and relatively smooth (except where they aren't as
you note there are singularities)... and you can usually afford throw
billions of cycles at them. While we weren't looking computers got
_fast_.
I don't think going with the easy route and just using blackbox optimizers
does the job best, here.
Absolutely not. But it gets you something right away.
Theorizing about the best psychoacoustic criteria doesn't— and it
seems like most of the people who have invested a lot of time thinking
about this problem go the closed source opaque route thus failing to
advance the public science... even many people who do proper research
in this space only output papers, which while informative often don't
do much to advance the _practice_ of surround listening (perhaps the
closed source folks implement their techniques).
That isn't being done now. Even to accelerate convergence, or to give a
global, smooth starting point for the optimization procedure(s), or to
regularize the eventual outcome. Why not? Are we really that lazy (well I
am, but are the researchers in the feel as lazy as me as well?)
Dunno about you, but I'm pretty darn lazy!
https://people.xiph.org/~greg/ambisonics/ambi_opt.c
Under xiph.org? Ooh! Please, more of that. And then more reseach plus
application in how to optimally code/decode even first order using Vorbis
(or some derivative?).
Back in 2007 I created a special Vorbis mode for coupled first order
b-format. I thought I posted about it here, but I can't find the
message with a quick search. I don't have a surround reproduction rig
so I was unable to listen to it, and I just wanted feedback about
how it sounded. IIRC I didn't get any feedback. ::shrugs:: I doubt I
still have the modes or files anymore.
My general belief is that (with carefully constructed formats and
encoders) fairly modest bitrate perceptually compressed audio can
sound excellent— and if we want to increase quality further with more
bits what we should be doing is going to (increasingly high fidelity)
surround before decreasing lossyness.Unfortunately, high quality
surround content and playback rigs are still scarce. ... and the
ITU/MPEG crowds approach to lossy surround coding appears to be
excessively parametric— very few bits but its mostly only good for
creating ping-pongy effects and motion sickness. (probably a catch 22
with the non-existence of good content and playback rigs, why make
lossy codecs for acoustic holography when people are only doing
quadrophonic-redux pingpongy stuff?)
In any case, the lack of practical deployment of really surround
systems has made surround coding stuff stay low on my priority list
even though I care about it personally— work in that area, for me, has
fairly little improve-the-world bang for the buck.
___
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound
