[LAU] Creating a personalised HRTF?

Fons Adriaensen fons at linuxaudio.org
Mon Jan 20 21:41:26 CET 2020

On Mon, Jan 20, 2020 at 04:20:29AM +0100, Jeanette C. wrote:

> is there any (convenient) way to create a personal HRTF (head related
> transfer function)?

Straight answer: no, repeat no. There are some commercial 
systems (Android apps) claiming to do this, using e.g.
pictures of your head and ears. AFAIK, none of them really
work, it's all hype and easy money.

> I've been wondering about this for a while. I knew only one way: go to a
> very tech savvy place - a uni maybe - and let them do it.

Indeed. And if these personalised HRTF have to be useful at all,
e.g. for binaural with head tracking, you need a lot of them,
in all possible directions.

I've been working on binaural reproduction for some years, and
at the same time some of my collegues have been exploring how
to get personal HRTFs, and how these could improve things.
None of the methods they tested has produced anything really
useful. In some of the listening tests we conducted, people have
evaluated their own personal HRTFs (obtained by different methods)
as not working at all, while some generic ones (from dummy heads)
worked very well.

There are other aspects of binaural reproduction that seem to be
much more important - can't go into the details, but room sound
and correct head movenent tracking are among them.

One reason is that your hearing will adapt to an HRTF set after
a few tens of minutes of listening, assuming the rendering system
is OK and provided the HRTFs uses are 'plausible', i.e. not
completely wrong. The adaptation is permanent if the exposure
has been long enough - you can come back later and things will
still work. This is only to be expected if you think of it for
a while. Your perception of reality doesn't break down if e.g.
you wear a cap because it's cold or raining, or when you move
your earflaps a bit forward or press them against you head.



More information about the Linux-audio-user mailing list