0
0

to Brett and FMOD team. Any news soon about the better than HRTF feature you mentioned a long time ago? Anything to do with this Nahimic 3d technology?

  • You must to post comments
0
0

[quote="John Schultz":1869lcnv]I’m nearing completion of another project right now, so I don’t quite have the time to experiment with FMOD’s DSP API, but off the top of my head, here’s how you could do it yourselves:

  1. Study FIR filters and FIR filter design using the FFT: there’s very good info on musicdsp.org and dsprelated.com.
  2. Obtain small microphones you can put inside your ears, such as the Radio Shack Optimus 33-3013.
    [/quote:1869lcnv]

My experience using impulses and convolution (it used the KEMAR data from MIT) for HRTF was that using ‘one pair of ears’ didn’t really work well…you would need to create a number of modeled ears…ideally each user would create the impulse of their own ears.

cheers,
Templar

  • You must to post comments
0
0

The company that was doing it for us has dissapeared, so in the meantime there is a fake hrtf that filters sounds when they are behind you, you can initialize this with FMOD_INIT_LOWPASS_OCCLUSION (this naming needs to be fixed)

  • You must to post comments
0
0

We are now working with a 3rd party to produce a newly researched HRTF system for FMOD that is ‘personalized’ to your head to give the best results. More information will be provided about this at a later date.

  • You must to post comments
0
0

Dissapointment sets in…

  • You must to post comments
0
0

Checked out the Nahimic website: fantastic claims and no sound samples?

From [url:1p7yzqz8]http://www.positive-feedback.com/Issue35/aes.htm[/url:1p7yzqz8]
[quote:1p7yzqz8]A company called A-Volute was selling a "3D Sound Expert" called Nahimic which they claim to produce surround sound and other fancy immersive imaging through a single pair of low-cost speakers. Their demo was not impressive.[/quote:1p7yzqz8]
Circa 1990, I had the opportunity to tour NASA Ames Research Center in Moutain View, CA. There I listened to Crystal River Engineering’s Convolvotron, which was custom hardware performing ~512MACs n-times/sec which produced a very high quality HRTF. We listened on Stax electrostatic headphones and our motion was tracked with 6-DOF Polhemus magnetic trackers. There was no per-person calibration: the quality blew me away. To date, I have not heard better simulated 3D audio. Perhaps google for Scott H. Foster and try to get in touch with him: his 3D audio work was awesome.

Binaural recordings also work well, even when not based on one’s own head. You can purchase a high-quality dummy mic head from Neumann
[url:1p7yzqz8]http://www.neumann.com/?lang=en&id=current_microphones&cid=ku100_description[/url:1p7yzqz8]

Directly interpolating FIR filters may not be optimal, however tricks can be used to minimize undesireable intermediate filter responses. For example, cross-fade/blend the post FIR filtered samples instead of interpolating the FIR filter coefficients (more expensive to compute).

An option to calibrate for one’s own head is cool, though this requires somewhat rare mics (most mics won’t fit in your ears); limited audience.

It should be possible to do very high quality 3D positional audio on modern PC’s in real-time, without custom calibrations. Scott Foster did it ~18 years ago ;-). Again, just adding a phase/inter-ear delay will dramatically improve spatial perception. I did this in 1993 on the Amiga: [url:1p7yzqz8]http://www.brightland.com/sourcePages/vr_slingshot.htm[/url:1p7yzqz8]. You could hear the guided missile flying around your head.

John

Re: Convolvotron: http://www.csun.edu/cod/conf/1993/proceedings/3d~1.htm
[quote:1p7yzqz8]In common amongst these is the notion that the operator is blind to the objects, be the blindness or the objects virtual or real. The Audio Image synthesis technique is based on the Head-Related Transfer Function (HRTF); the listener-specific, direction-dependent acoustic effects imposed on an incoming signal by the outer ears. HRTFs in the form of Finite Impulse Responses (FIRs) are measured with small probe microphones placed near the two eardrums of a listener seated in an anechoic chamber for 144 different speaker locations at intervals of 15 degrees azimuth and 18 degrees elevation. In order to synthesize localized sounds, a map of listener-specific "location filters" is constructed from the 144 pairs of Finite Impulse Response (FIR) filters represented in the time domain. The map of FIR filters is downloaded from a computer to the dual-port memory of a real-time digital signal-processor, the "Convolvotron," (ref.). The device convolves an analog signal with filter coefficients determined by the coordinates of the target location and the position of the listeners head, thus "placing" the signal in the perceptual three-D space of the user. The current four-convolvatron configuration allows up to 16 independent and simultaneous sources and is capable of more than 1200 million multiply-accumulates per second. The resulting data stream is converted to left and right analogue signals and presented over headphones.[/quote:1p7yzqz8]

Re: Stax: audiophiles, see: http://www.yamasinc.com/srs-007mk2.htm , http://www.stax.co.jp/index-E.html

  • You must to post comments
0
0

latest release now includes it as its own flag, FMOD_INIT_SOFTWARE_HRTF

  • You must to post comments
0
0

[quote="John Schultz":1hmtospb]Checked out the Nahimic website: fantastic claims and no sound samples?

From [url:1hmtospb]http://www.positive-feedback.com/Issue35/aes.htm[/url:1hmtospb]
[quote:1hmtospb]A company called A-Volute was selling a "3D Sound Expert" called Nahimic which they claim to produce surround sound and other fancy immersive imaging through a single pair of low-cost speakers. Their demo was not impressive.[/quote:1hmtospb]
Circa 1990, I had the opportunity to tour NASA Ames Research Center in Moutain View, CA. There I listened to Crystal River Engineering’s Convolvotron, which was custom hardware performing ~512MACs n-times/sec which produced a very high quality HRTF. We listened on Stax electrostatic headphones and our motion was tracked with 6-DOF Polhemus magnetic trackers. There was no per-person calibration: the quality blew me away. To date, I have not heard better simulated 3D audio. Perhaps google for Scott H. Foster and try to get in touch with him: his 3D audio work was awesome.

Binaural recordings also work well, even when not based on one’s own head. You can purchase a high-quality dummy mic head from Neumann
[url:1hmtospb]http://www.neumann.com/?lang=en&id=current_microphones&cid=ku100_description[/url:1hmtospb]

Directly interpolating FIR filters may not be optimal, however tricks can be used to minimize undesireable intermediate filter responses. For example, cross-fade/blend the post FIR filtered samples instead of interpolating the FIR filter coefficients (more expensive to compute).

An option to calibrate for one’s own head is cool, though this requires somewhat rare mics (most mics won’t fit in your ears); limited audience.

It should be possible to do very high quality 3D positional audio on modern PC’s in real-time, without custom calibrations. Scott Foster did it ~18 years ago ;-). Again, just adding a phase/inter-ear delay will dramatically improve spatial perception. I did this in 1993 on the Amiga: [url:1hmtospb]http://www.brightland.com/sourcePages/vr_slingshot.htm[/url:1hmtospb]. You could hear the guided missile flying around your head.

John

Re: Convolvotron: http://www.csun.edu/cod/conf/1993/proceedings/3d~1.htm
[quote:1hmtospb]In common amongst these is the notion that the operator is blind to the objects, be the blindness or the objects virtual or real. The Audio Image synthesis technique is based on the Head-Related Transfer Function (HRTF); the listener-specific, direction-dependent acoustic effects imposed on an incoming signal by the outer ears. HRTFs in the form of Finite Impulse Responses (FIRs) are measured with small probe microphones placed near the two eardrums of a listener seated in an anechoic chamber for 144 different speaker locations at intervals of 15 degrees azimuth and 18 degrees elevation. In order to synthesize localized sounds, a map of listener-specific "location filters" is constructed from the 144 pairs of Finite Impulse Response (FIR) filters represented in the time domain. The map of FIR filters is downloaded from a computer to the dual-port memory of a real-time digital signal-processor, the "Convolvotron," (ref.). The device convolves an analog signal with filter coefficients determined by the coordinates of the target location and the position of the listeners head, thus "placing" the signal in the perceptual three-D space of the user. The current four-convolvatron configuration allows up to 16 independent and simultaneous sources and is capable of more than 1200 million multiply-accumulates per second. The resulting data stream is converted to left and right analogue signals and presented over headphones.[/quote:1hmtospb]

Re: Stax: audiophiles, see: http://www.yamasinc.com/srs-007mk2.htm , http://www.stax.co.jp/index-E.html[/quote:1hmtospb]

Thanks for the links to these resources John! As Brett has mentioned, getting high quality 3D audio in headphones is on the agenda.

cheers,
Templar

  • You must to post comments
0
0

So are you guys going to be doing your own HRTF model or are you going to try and outsource it?
ETA?

  • You must to post comments
0
0

I know that htis is an old topic, but I am very interested to see if FMOD and personal audio are still going to, or have already , teamed up.

Also, how and where is the HRTF functions handled? Is it a built in algorithm within the FMOD system?

is there an ability to control the HRTF function, or would you design your 3D events and parameters and if someone uses headphones, they would just hear the positional cues translated into the HRTF function?

If I am asking the question wrong, but you know pretty much what I mean, please do your best to decipher on your end.

😛

  • You must to post comments
0
0

There is a fake hrtf already in, as i just said, it uses FMOD_INIT_SOFTWARE_HRTF

  • You must to post comments
0
0

[quote:eh9phfmu]I am very interested to see if FMOD and personal audio are still going to, or have already , teamed up. [/quote:eh9phfmu]
Yes, we are working closely with Personal Audio for an immersive 3D sound experience using headphones.

http://www.personalaudio.com.au/media.pdf

[quote:eh9phfmu]is there an ability to control the HRTF function, or would you design your 3D events and parameters and if someone uses headphones, they would just hear the positional cues translated into the HRTF function?[/quote:eh9phfmu]
We cannot give out these kinds of implementation details at this point.

Sorry to be so vague but that is the nature of the beast. Suffice to say this is very exciting technology, the PA guys have done an excellent job.

-Pete

  • You must to post comments
0
0

Yes, I read that. My question was regarding future plans for true HRTF.

  • You must to post comments
0
0

Is there any news on this?

If not, I’m developing my own HRTF system using a custom dsp.

Anyone doing this? Anyone already done this before? :)

  • You must to post comments
0
0

no there are no plans.

  • You must to post comments
0
0

Hangs head in dissapointment…

  • You must to post comments
0
0

Hi,

I’m also interrested to have a better HRTF function in FMOD. Of course, implementing this is a real challange, but it would also be a unique FMOD feature.
I’m working on a 3d audio-only game for blind people in which I would like to have a realistic environment rendering. And for now, a simple lowpass filter is not sufficient.
I do think I’m not the only one interrested in this feature. Not everyone can have a good 5.1 sound system at home.

Regards,

  • You must to post comments
0
0

I’m nearing completion of another project right now, so I don’t quite have the time to experiment with FMOD’s DSP API, but off the top of my head, here’s how you could do it yourselves:

  1. Study FIR filters and FIR filter design using the FFT: there’s very good info on musicdsp.org and dsprelated.com.
  2. Obtain small microphones you can put inside your ears, such as the Radio Shack Optimus 33-3013.
  3. Place a single speaker in a relatively echo-free room and record sine waves at many frequencies (higher frequencies will be more relevant) using both microphones placed in both ears. Do this for 8 or more points on a circle around your head.
  4. You can compute the inter-ear delay (effectively phase) at each point using cross-correlation of the stereo recording: you can easily convert the sample offset into milliseconds. Just adding an inter-ear delay or phase control will dramatically improve perceived 3D spatialization.
  5. Perform the FFT on the left and right recordings. Since you know the original frequency, you can examine the spectral components from the FFT to determine the amount of filtering for the head-to-speaker position. You can first record an open mic in front of the speaker to get a baseline recording.
  6. Create a FIR filter with the desired filter response for many different frequencies at each position on the circle, for each ear. You’ll have to experiment here, and may also be able to combine many different known frequencies (Fourier Synthesis) to save time on the analysis phase.
  7. With 8 sample points, you’ll have 16 FIR filters (8 per ear). At run time, simply interpolate the FIR filter coefficients based on position around the circle. Apply the resulting FIR filter to the audio stream (fast and easy, just a series of multiplies and adds), and you have a very fast, simple HRTF for a circle in the plane around your head. Use the shortest FIR filter length possible to reduce CPU load and reduce delay effects.
  8. Once you are satisfied with the results, capture 6 more points going around your head in a circle from front, to top front, etc. Then capture 4 more going around the final axis (you’ll already have the other points). Thus, 18 points on a sphere (stereo: 36 recordings). Perhaps read about “normal quantization” to pick the neareast 3 points to interpolate coefficients, or perhaps make a sphere of triangles, precompute the weight values for each vertex, intersect a ray with the correct triangle, then use barycentric interpolation to get the weights for the 3 FIR filters to interpolate at runtime.
  9. I’m currently doing all my real-time audio work in floating-point: it’s plenty fast on today’s PCs (web app running on millions of PCs). If FMOD’s DSP API supplies floating point samples, use them directly, otherwise convert them to/from floating point (use asm fistp to convert float => int). Dithering and noise-shaping aren’t really needed for a game application.
  10. Keep in mind, the HRTF is useful when listening with headphones. Some changes will be required for an open two-speaker system.

A custom HRTF DSP filter can easily run in real-time with minimal CPU load using FMOD…

  • You must to post comments
Showing 17 results
Your Answer

Please first to submit.