Politis Archontis wrote:
Hi Jörn,
On 03 Jun 2016, at 15:27, Jörn Nettingsmeier
<netti...@stackingdwarves.net<mailto:netti...@stackingdwarves.net>> wrote:
Note however that while the quality of first-order to binaural is quite good
because the listener is by definition always in the sweet spot, first-order
over speakers can be difficult for multiple listeners when they're far outside
the center.
This is by no means meant to provoke, but I have never managed to hear a
convincing B-format to binaural rendering, or to produce one myself. Could you
possibly share some info on the decoding approach that you used that results in
a good example?
In my experience, no matter how much tweaking in the decoding, there is severe
localization blur, due to the large inherent spreading of directional sounds,
and low envelopment due to the wrong (high) coherence in the binaural signals
with reverberant sound, compared to the actual binaural coherence. And there is
also serious colouration, with a loss of high-frequencies, that seems
direction-dependent. The fact that everything is on the ideal sweet spot under
free-field conditions doesn’t seem to improve much, it actually seems to do
more harm (I believe that a small amount of natural added decorrelation from a
room and tiny misalignments from speakers etc. seem to improve binaural
coherence and the perceptual quality somewhat of loudspeaker B-format
reproduction).
As long as we we don't use personalised HRTFs (accessible HRTF
personalisation methods already seem to be in development), objective
comaprison between loudspeaker and binaural decoders are just not possible.
Otherwise, I don't see any real reason why binaural decoders could not
introduce some of the desirable decorrelation and coherence effects you
explain.
Listening to a binaural rendering from a real B-format recording is not so bad, there is no reference for comparison, but for VR the difference I’ve heard between using directly HRTFs and B-format rendering is huge. And as many of these applications rely on sharp directional rendering with accurate localization of multiple sound events, traditional B-format decoding seems unsuitable to me.
This is not really disputable.
However, you can still mix an FOA recorded "location background" and
(computer generated) object audio.
Kind regards,
Stefan
The performance improves somewhat with 2nd-order rendering, and significantly
with 3rd and 4th-order rendering. Also it improves dramatically using plain
B-format with a well-implemented parametric active decoder, such as HARPEX or
DirAC.
So my guidelines for VR till now have been,
a) if bandwidth is not an issue go for HOA rendering (at least 3rd-order),
b) if it is an issue, like the streaming application of the OP, stream B-format
and use an active decoder at the client side.
But I’d like to hear many opinions on this too, and any counter examples!
Regards,
Archontis Politis
_______________________________________________
Sursound mailing list
Sursound@music.vt.edu
https://mail.music.vt.edu/mailman/listinfo/sursound - unsubscribe here, edit
account or options, view archives and so on.
