Binaural sound generation.

February 19, 2005 1:48 AM

I'm looking for some algorithms for generating binaural sounds.

I'm attempting to create binaural sounds for people who use headphones, read: sounds that allow stereo headphones users to exactly position where the sound is coming from (up/down/front/back in addition to left/right). The environment of these sounds is a hypothetical infinite empty space with sea-level air pressure so the traditional surface modeling issues can be completely ignored. While some parts of the exercise are obvious (delaying sound playback in either ear based on the position of the source, distance between the virtual ears, and the speed of sound), other parts are not (modelling frequency shifts from ear shape and the auditory canal).

I see no method of distingushing front from back nor up from down without acoustic modelling of the ear itself. Google has failed me completely, so any algorithms for this would be greatly appreciated. Bonus points for any algorithms that can be kludged into a close-enough approximation for realtime.

posted by Ryvar to science & nature (10 comments total)

I should mention I've heard a lot of stuff done with EAX/EAX2/A3D and it was pretty universally horrible compared to custom sounds created through recordings created by inserting miniature microphones into the recording party's ears. I'd like to generate sounds of the quality found in the latter method if at all possible.
posted by Ryvar at 1:56 AM on February 19, 2005

Hello.

I am interested in sound art, so only dip into technical stuff.

I created a soundtrack that had the illusion of rising height. The method is a very crude idea of a Head Related Transfer Function.

Create 2 copies of a sound. Parametric EQ one of them with a notch at 6khz. As you crossfade the signals to the EQd version, it creates the feeling of rising height. You need to play with the Q of the filter and the exact position of the notch to get it working.

This is quite good at making a sweep upwards or downwards, but I doubt if it will work well for positioning one off sounds.

It's not the answer you're looking for, but hope it is of some interest.

What is the project?
posted by lunkfish at 2:30 AM on February 19, 2005

That's certainly more helpful than what Google's turning up, so thanks.

BTW, for an example of the effect I'm going for, put on some headphones and listen to the sound effect on this page.

I have yet to hear EAX/EAX2/A3D do something even remotely that good.

As for what the project is - I've done a bunch of OpenGL/DirectX tech demos in my time, but I'm only just now becoming interested in sound, and I wanted to try to create a realtime system with approximated binaural output for headphones users. I'd eventually want to look into surface modelling (way) down the road, but for now I'd like to nail down just the basics of a sound circling the user in all three spatial planes.
posted by Ryvar at 2:52 AM on February 19, 2005

Some equations in here: Spatial Hearing Mechanisms and Sound Reproduction.
posted by AlexReynolds at 3:00 AM on February 19, 2005

This looks an awful lot like what I was looking for. If anybody has some good additional links, I'm all ears (no pun intended).
posted by Ryvar at 3:01 AM on February 19, 2005

Ambisonic bpan VST plug-in code.
posted by AlexReynolds at 3:05 AM on February 19, 2005

You may want to look into the literature of assistive technology.

I once worked with a researcher who told me the best 3D sound recording he ever achieved was by placing a set of minirecorders in the hollow head of a mannequin. You have no idea how seldom this tidbit comes up in casual conversation. If you're interested in learning more, please email me; I should have the article archived somewhere. The illustrations are fairly nifty.
posted by melissa may at 3:10 AM on February 19, 2005

Shoot, Ryvar, I reread your question and it looks like I was giving you advice on exactly what you didn't want to do. Oh well. I do like using the phrase "minirecorders in the hollow head of a mannequin" every chance I get.

Way to QED it up, AlexReynolds.
posted by melissa may at 3:19 AM on February 19, 2005

'liss: minirecorders in the hollow head of a mannequin is more or less how the pros do it too!

eg. http://www.neumann.com/infopool/mics/produkte.php?ProdID=ku100
posted by BobInce at 9:15 AM on February 19, 2005

While the VST plugin is the coolest .... mannequin mounting isn't necessarily the ultimate in physical modeling. Some people (mostly those recording live shows) use things like these mounted on real heads (usually their own).
posted by weston at 12:33 PM on February 19, 2005