Improve front / back projection of binaural renderer ?

[M Dammer]

I just started experimenting with SSR with the standard files that come with the current git clone. I am pretty impressed by this project but I wonder if the front / back "projection" of the binaural renderer can be improved. In the tests I did with different pairs of headphones putting a sound source in front of me always resulted in the sound appearing somewhere close or in front of my brow or sources behind the head seem to appear at the base of my skull but not really behind or before me. Just for clarification: I have a fairly "normal" head and average hearing. The right / left in contrast sounds excellent and does appear outside the head as it should be.

I would appreciate any help about how binaural rendering can be improved. It looks as if the latest version from github supports SOFA files that are meant to contain a full 3D HRIR.

However I do not know what files to use and what command line arguments to use in order to use them.

Thanks.

[Jens Ahrens]

Hi Mark,

Great that you're using SSR!

The aspect of front-rear ambiguity in binaural rendering has many layers, most notably the individuality of HRTFs (even if your head and ears are as average as they can be...) and the presence or absence of head tracking (i.e., whether or not the rendering system reacts to your head rotations and makes the sound source positions appear static in the space around rather than relative to your instantaneous look direction). In short, head tracking will resolve a lot of the ambiguity. But you'll need a head tracker (i.e., a motion sensor) for that that needs to be attached to the headphones. Here is one that SSR supports: https://github.com/Razor-AHRS/razor-9dof-ahrs/wiki/Tutorial I think that the parts are unfortunately not available anymore. We haven't created new designs because we still have enough copies of this old one lying around.

Other than that, here's an attempt to improve the situation in static (i.e., non-head-tracked rendering): https://www.researchgate.net/publication/325392160_Simple_Reduction_of_Front-Back_Confusion_in_Static_Binaural_Rendering

The authors essentially emphasize the acoustic differences between HRTFs for front and rear directions. You might want to try that with the HRTFs that we ship with SSR (https://github.com/SoundScapeRenderer/ssr/tree/master/data/impulse_responses/hrirs). But there is, of course, no guarantee that this will work for everybody / every set of HRTFs.

BTW, SSR will eventually support 3D sound scenes at some undefined point in the future. The SOFA interface is one of the bits that is completed. But it won't have an effect on the front/rear thingy, though.

I hope that this helps!

Best regards,

Jens

--

http://www.ta.chalmers.se/people/jens-ahrens/

--
You received this message because you are subscribed to the Google Groups "SoundScape Renderer" group.
To unsubscribe from this group and stop receiving emails from it, send an email to SoundScapeRende...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/SoundScapeRenderer/e932ec51-15f8-4baa-b3bb-2f674e91d9c6%40googlegroups.com.

[Matthias Geier]

On Fri, Jan 31, 2020 at 4:39 PM 'M Dammer' via SoundScape Renderer
<SoundScap...@googlegroups.com> wrote:
>
> [...]

>
> I would appreciate any help about how binaural rendering can be improved. It looks as if the latest version from github supports SOFA files that are meant to contain a full 3D HRIR.

As Jens already mentioned, this doesn't really change anything
regarding front/back confusion in general.

But with SOFA support, you get access to a lot more data sets, which
may or may not be better suited for your head.

> However I do not know what files to use and what command line arguments to use in order to use them.

I can't really tell you what files to use, you'll have to try some and
hear if they work well for you or not.

But there's lot (probably too many!):

A good overview page is this:
https://www.sofaconventions.org/mediawiki/index.php/Files

I recently tried the data sets from
http://audiogroup.web.th-koeln.de/ku100nfhrir.html, which provide
multiple distances somewhat close to the head.

As for how to use them with the SSR:
This is the same as before, you just have to specify a .sofa file
instead of a .wav file.
For example:

ssr-binaural --hrirs my-favorite-hrirs.sofa

Alternatively, you can specify HRIR_FILE_NAME in the SSR config file.

If you have further questions (or any problems), don't hesitate to ask!

cheers,
Matthias