On 2022-12-30, brian.k...@sorbonne-universite.fr wrote:
I will chime in here on this discussion again.
As will I, as a pure amateur.
While HRTF selection/individualization is a work in progress, the use of selection combined with ITD individualization and better yet with headtracking tackles many of the issues. I do not think personal ear molds and the like are required, though they are a nice way to get individual HRTFs via gestalt measurements (either on a dummy head or numerical simulation).
Ear molds are not necessary, indeed. They only go as far as the ear canal, which is a *very* small part of the overall HRTF/IRTF. Just by physical measurement, it affects only the very highest frequencies, much of which people of my age (43a) can't even hear.
What is more to the point is the shape of the pinnae, the size of the head (separation between the ears across the skull), and the upper torso. Also, how much fat you have there makes a difference in the lower frequencies.
And in fact, one of the things nobody speaks of is how we *too* actively listen to sound. We might not have *quite* the ears of a bunny, but we *do* possess muscles which orient our ears dynamically towards an oncoming sound. I for instance learned to consciously control some of those muscles by example of my second grade teacher; I can still pull back my earlobes a centimetre or so by command, and I'm acutely aware of them doing their thing otherwise when I'm locating point sources under acute stress.
I believe these sorts of things which haven't been well accounted for, are the reason binaural fails, over and over. Despite its theoretically high potential. Why headtracking fares so well, and why headtracking combined with ear moulds doesn't quite compound as we'd expect.
Head dimensions provide a very good indication of ITD variations, and can be used to individualize HRTF data on that front.
Yes. Also neck and shoulders, height above ground, torsion, because those lead to combing from the whole torso and ground up. Dynamically even. Though that is then very difficult to measure and simulate.
I know of no studies that should there are significant individual HRTF differences resulting from fat content or torso shape.
(You ahead of me.) There may not be studies, but basic physical acoustics tells us they will be there. They will be indiscernible beyond the middle range of say 1-2kHz, but below, they will do stuff. Because now you'll have body resonances and then bone conduction to the inner ear.
Hair/head covering is a different factor, and has been shown to be audible.
Of course it is. It's a highly nonlinear and dissipative medium, and even in the regime of it just being heavy in aggregate, obviously it dampens down what we hear.
But then it's also stuff our neural circuits have evolved and learned to tune out. It's a distraction, *well* tuned out. (I keep a ponytail. Even if it sometimes flares out, rarely does that take out my directional hearing for more than a few seconds.)
However, the effect would not be expected to beyond understanding, and one could imagine spatial correction filters to modify the HRTF relative to such absorption characteristics.
Do we need adaptation here at all? I'd argue relying on neural adaptation on part of the subject would fare much better.
It all remains a linear system and such layers could be separately processed, as with ITD processing.
There's nothing linear about this. And I'd argue, there shouldn't be. Linear is maybe easy, but then trying to keep it all linear, isn't how it should be done. In some cases, going with the nonlinearity of the basic problem is *easier* than trying to linearize it.
It must be repeated that our auditory system adapts to our own local changes, in clothing, hair style, etc. and we are not significantly thrown off by such things (at least after adaptive listening for a bit). So, such changes can be feasibly incorporated, with the inclusion adaptation.
FFS, great minds seem to think alike. Again, you're ahead of me. :)
As for the multichannel headphone proposal, this does not alleviate all the above non-pinna variations mentioned.
Also, it's *very* difficult to reproduce multiple moments within a small headphone capsule. That's because you cannot emanate soundwaves pointwise, using any kind of compact actuator. In ambisonics, you *can* emanate near-planewaves, and in NFC-HOA or WFS, you *can* put in constructive wavefronts, from afar. But at headphone distances, that is almost impossible. It is almost impossibly unstable to do as well.
However, it is an interesting notion to provide wavefronts tot the indvidual's actual pinnae (requiring only multichannel head/torso related intensity vector transfer functions) that has been developed on and off over the years.
Yes. So I'd argue it has to be done from afar. First, via Ambisonics to get our ears to central low order thingies, and then towards extended area holophony. Because this sort of thing utilizes the normal directional hearing of a human being to its max. It doesn't feign to trick it like imperfect binaural does, but just tries to reproduce a progressively better soundfield, to be listened to by your extant ears, shoulders, big head, and whatever girth you have.
The main issue in my experience has been the acoustics within any type of headphone cavity which make creating directional wavefronts almost impossible. For example, the prototype of Greff used an open grid with speakers, providing interesting results, but once enclosed for better frequency bandwidth and a commercial device the same results were not achieved.
Also take my amateurish analysis. You see, even if you somehow can reproduce a soundfield inside the headphones, how the fuck did it interact with the pinnae and the torso? As the short HRTF already said it should? "Head related transfer function".
Well it didn't. So it sounds sterile. And if you even tried to do *that* your earphones would have to be rather huge, multiply active, ... No.
-- Sampo Syreeni, aka decoy - de...@iki.fi, http://decoy.iki.fi/front +358-40-3751464, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2 _______________________________________________ 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.
