I recently became curious about the current state of the art in headphone correction and HRTF measurement and have been trying out 3D audio software. As part of this I have tried to find a cheap real ear measurement system similar to those used for fitting hearing aids but was unsuccessful. The Mini DSP system doesn’t have the correct adapters. I wanted a personal linear phase measurement system for headphone correction and playing with HRTF filters.
If I was a premium headphone company I would consider offering this service to my customers just like all the Audiologists on the high street do for hearing aid fitment.
Does anyone know if this is happening as it is quite easy to set up with a PC a couple of ADCs and DACs and some speakers and some probe tubing on a microphone? You would have to use MATLAB or similar as a one off but a simple app would be usable by non tech types in a shop or at home for custom fitment. It could even be built in to some IEMs.There is also the option on using an in situ audiometry target - can see no sign of that anywhere at the moment?
Those articles look encouraging. I think there will be an occlusion effect with the in ear microphone though and additional resonance with the technique shown. In general a thin tube probe tip is required to go into the deep canal within 5-7mm of the tympanic membrane to get the true levels based on current guidance for hearing aid fitment. Otherwise it will be out by a few dB. In general you would use a pseudo random spread spectrum sound input from a speaker in a sound treated room and then take the FFT of the epoch averaged canal probe output and compare against the target, the aid is reprogrammed so the output matches the target and output is rechecked. The probe is calibrated against a monitor microphone which which samples simultaneously outside the ear. With a bit of extra processing you can also get a linear phase transfer function out of this system for a HRTF or accurate headphone correction although hearing aid targets just use scalar rms amplitude.
You can buy all this stuff but it is tens of thousands of dollars for a turnkey system and it is designed to program hearing aids against a prescribed fitment target based on the subjects audiogram not a headphone target - it would need to output the predicted headphone target FIR filter to feed the Convolver in your audio software or hardware and I don’t think the current real ear systems will do that or even put out the raw data for you to process yourself.
Maybe the paper describes something similar, I will get my hands on it and have a look.
I bet someone is doing this on the cheap somewhere.
This looks like a good start, not using standard probe tube microphones still but that is just a detail. I will have a good look, its being done slightly differently to how I had expected but amounts to the same sort of thing I think. The room ambience is done via sampling the actual room by the looks of it. Can’t see any head tracking though which you do need for a realistic experience.
I’m very interested in your results but you should probably also read the audiosciencereview.com link that @cpp posted completely if you haven’t already. There are definitely some issues with the accuracy of measuring frequency response using this type of relatively cheap in-ear microphones. Two things that come to mind are that the microphones are not calibrated and that they will not have the proper acoustic impedance.
I don’t think the Impulcifer project calibrates the microphone but that asr.com thread suggests some ways you could do so. As for acoustic impedance, it may not be a big issue as long as you are using headphones with a flat enough frequency response.
Also take all of the above with a grain of salt because I am not a professional, nor do I play one on TV.
I’ll bet that’s because you’d have to make many measurements at various angles and then interpolate them somehow during playback as you turn your head.
Had a look at the AES thread. They seem a bit disconnected from current practice in audiology for example https://www.aes.org/images/e-lib/thumbnails/3/9/3994_full.png deprecates real ear measurement but as I said it is on every Main Street at least in the UK for hearing aid fitment and it is in all the national guidance - https://www.audiologyonline.com/articles/real-ear-measurement-basic-terminology-1229
There is a lot of talk about personal transfer functions - you can spend $500 on a predicted HRTF based on a video of your head for example. I think a good start would just be accurate linear phase headphone correction then the 3D aspects would come later. I think you could pop down to the local audiologist with your headphones to have it done then load it up and off you go.
This is interesting - it uses in situ audiometry as I suggested at the
start of the topic. There are preset profiles for different headphones and then additional personal preference eq which they call “Sound ID”.
There is no real ear measurement but in theory this is less of a issue if in in situ audiometry is used to help predict a headphone correction filter using reference headphone model data as a baseline. It won’t be linear phase though and you can’t use it for HRTF.
It’s pretty cool but even Sonarworks states the hearing test feature is not a very accurate measurement of your hearing. I got subjectively much better results using headphone correction EQ from oratory1990: https://www.reddit.com/r/oratory1990/wiki/index
You should consider posting your thoughts on real ear measurements on his Reddit community as he knows far more about accurately measuring headphone frequency response: https://www.reddit.com/r/oratory1990/
Surprising that this isn’t built into more vr media players, some games have it, I’m still using waves NX 3D audio driver app for windows which I like and you don’t need a vr rig. That costs about the same if you buy the Bluetooth headphone tracker.
Will have a look on the Reddit for headphone correction.
It’s surprising how many different groups are doing the same sort of stuff. Wouldn’t be surprised if Apple or google make it part of the OS or the AirPods (they already have head tracking and 3D audio but don’t support stereo media at the moment only 5.1 encoded or better.)
I think to get it right, a person needs to go where Audiologist go and that is using a screening audiometers , diagnostic audiometers and/or clinical audiometers. or Tympanometry machines. But of course none of us are Audiologist or ENT’s. My cousin is an MD in Otorhinolaryngology at Mayo and notes, you would be surprised at the number of patients they see that have damaged their hearing by loud music, loud noise or putting things in their ears or loud hearing test. Me I will just leave it up to the pros’. My hearing is just to delicate to screw up at my age. Be safe and careful.
Unfortunately the commercial real ear measurement systems I have looked at only go up to 10kHz and only output RMS amplitude transfer functions for target matching, Sadly they don’t seem to have accessible APIs for use with headphone correction either. This will have to go on the back burner as a project for me for the moment.
Sorry I wasn’t clear, I did have a look at the post before - the impulcifier project looks to be using standard external canal microphones and on a practical point it doesn’t seem to have head tracking either which I find limits 3D imaging for me, when I turn off head tracking I lose the sense of being in the studio with the musicians.
It appears that nobody is using real ear measurement with a probe microphone tube positioned next to the tympanic membrane for headphone correction/HRTF at the moment. This isn’t easy as you have to get close to the drum to avoid the canal antinode especially at higher frequencies. Also no sign of really accurate in situ audiometry as an alternative. That does not mean that the existing technology doesn’t work quite well of course or that these other options are necessarily practical or better in the real world. There are theoretical reasons why they might give a more accurate measurement of the overall system transfer function though. I think this falls into the research domain at the moment and although I have some experience in this area I don’t have the time to pursue it myself just now.
I was looking for an appropriate spot to post one of my notorious semi-serious questions that is probably an out-of-date can of worms. Attention @generic, I know you like to comment on these, so I’m alerting you. (hah - will that make you want to not comment? or will this parenthetic remark?)
So in the soundstage discussion, we talked about FR and headphone design and did mention HRTF.
Recently, a change of medication has been helping me lose weight. (Mounjaro, not Ozempic). Weight that in part I’d put on due to other meds that had been prescribed for diabetes.
HRTF… I notice that my face has been getting slimmer. And then I started thinking about fat people and thin people. And young and old (ears never stop growing).
Is there any research that has been done to know if weight gain or loss affects the length or shape of the ear canal, or other physiological aspects of HRTF?
And here is the amusing can of worms. Should audiophiles feast or famine?
