[RRod]

I've been playing around with a couple of HRTF databases (IRCAM and CIPIC) to try and get some movie/music 5.1 files into a semi-believable virtual headphone surround. So far I haven't been impressed with the results. Is there a principled way of going about this?
 
So far I've been using stv014 convolve code to apply the impulse responses to the isolated 5.1 track files, but I'm sure this is probably too simplistic a solution. I've mainly played with the IRCAM files, which have "raw" and "compensated" tracks (having something to do with diffuse field), and have been using a couple of dummy head responses that I've found work well with my OpenAL setup. But I'm finding the discrete sounds in the games work much better than the 5.1 tracks.
 
Any thoughts appreciated!

 

[stv014]

I think virtual surround solutions like Dolby Headphone also add some reverberation to simulate room acoustics. Perhaps that could improve the effect by providing additional cues from the early reflections. Or they just used better HRTF impulse responses. In any case, the approach of convolving each channel of the 5.1 input with a stereo IR and then mixing a stereo output from that should work in theory, the challenging part is creating a set of good impulse responses. Head tracking would be even better, but that is obviously more complex and expensive to implement.
 
Quote:

Originally Posted by RRod /img/forum/go_quote.gif
 
I've mainly played with the IRCAM files, which have "raw" and "compensated" tracks (having something to do with diffuse field)

 
That probably means an EQ was applied that makes the overall frequency response of the HRTFs (either the entire set, or a subset like frontal or horizontal only) flat.

 

[RRod]

   
That probably means an EQ was applied that makes the overall frequency response of the HRTFs (either the entire set, or a subset like frontal or horizontal only) flat.

 
Sounds likely given the description here:
http://recherche.ircam.fr/equipes/salles/listen/download.html
 
Applying the compensated samples sure does sound more natural, at least.
 
Good to know I wasn't horribly off in my approach. I started with the IRCAM samples b/c I have used them before with OpenAL, and thus already had a dummy that worked for me in that setting. Next step is to see if I can get better results with the other samples, hopefully uses a less obnoxious test sound than the IRCAM site provides ^_^