Originally Posted by GuySonic
Great sounding binaural recordings guys! Appreciate a thread sharing actual recordings along with the thoughtful discussion. There's a common misconception about most all manufactured binaural in-ear dummy head microphones some of you own or have at times used to make binaural recordings. Now from what I’ve read, those designing in-ear dummy head products first tried developing a working product with direct microphones output, but found some inherent problems no mater what ear shape was used. As discussed, no ones ear shape is identical, so a dummy would solely be its own prefect customer for hearing these recordings. However, more-or-less everyone else listening on phones would not be so thrilled. So to have a practical dummy head product appealing to a wider audience special filter networks were designed into these dummy head microphones to greatly modify the output signal. In other words, these dummy head products are not your father’s (or your own in-ear) binaural having a pure unmodified signal into the recorder. Instead these heads produce a highly modified-in-ear-semi-binaural appealing to most listeners wearing phones, and work better (if for only a short listening period) as resource special effects ‘mixed-into’ into commercial releases meant for both headphone and speaker playback. While these made-for-general-consumption dummy heads are popular with most binaural recordists, they are a far cry from true personally worn in-ear mics both for using output network filters, AND for having completely different acoustic response. These commercially available dummy head mics use many materials differing greatly from our own water-based-flesh-and-bone acoustic characteristics, and as such, might well be classified as ‘effects’ mics producing a baffled mic output too greatly differing from the natural HRTF response to be even considered HRTF. Yes, they can sometimes produce nice sounding recordings, but these while sounding OK for nature sounds for which we have little reference for how they should sound, are too highly colored sounding, have quirky imaging, and generally too inconsistent for acoustic musical instrument recording purposes and not yet, or likely in the future, be considered appropriate as studio mics. For these reasons, both from accidental discovery and intentional acoustic research, I have developed and patented (recently expired) a different method of producing recordings without modifying the mic’s output signal or changing the correct water-based-flesh human acoustic head response for true HRTF recording ability. The DSM (dimensional stereo-surround) mics I make and use are modified physical/acoustically to be neutral (uncolored), and may be personally headworn or placed on an acoustically correct HRTF baffle (the only true HRTF one available) I also custom produce. So while we are sharing our recordings made with our mics, suggest not being too narrowed about your own fine recordings to listen/comment to some of mine done over a 25 year period by myself and others using the not-in-ear-DSM microphone approach. In other words, if you like binaural, and want to see this a virtual reality recording method used in more commercial stuff, consider a slightly more practical and more effective way to getting true you-are-there-audio into commercial music products. You already have enjoyed the sounds recorded by DSM mics in many, many, many major films (like Lord of the Ring Trilogy, Sin City, and others going way back to early 90’s). But the music professionals/industry experts have their heads stuck in other places, and are NOT going to let go of what they think they know works profitably for them until proven otherwise by a major player successfully using this old but still new technology. So suggest we keep our passion for this type of sound, it will eventually be accepted by the music industry; maybe even in our own lifetimes! Here’s just a quick musical sample: http://22.214.171.124/mp3/gkbjplat.mp3 Much more musical stuff at: http://www.sonicstudios.com/mp3_2slp.htm
Well, first things first. The recordings that I heard were very, very good. Kudos to you on that and what you have accomoplished.
As far as the other issues that you raised with regard to the accuracy of the HRTFs employed in such mannequin heads as KEMAR, Head Acoustics, Neumann, Bruel & Kjaer et al, I'm not sure what to say to convince you to change your opinion. I know as a matter of fact that such heads are the standards used in research...by pretty much every university and pretty much every product development team concerned with product sound quality (whether in automotive, aviation, pedestrian signaling devices, the appliance industry et al) uses them.
All of these activities require one thing at their core - extremely-realistic recordings of sounds suitable for playback for critical listening tests and evaluations; as such extremely spatially-accurate recordings are required. This is particularly true in the automotive world where the vehicle interior acoustics are very complex (regions tending toward diffuse, regions tending towards absorptive, near field effects etc); listeners consistently judge their experience in listening to binaural mannequin head recordings made in the vehicle
with what they actually hear while in the vehicle. Anyway, this level of spatial and temporal accuraye is required because juried listening tests are regularly executed using binaural recordings (made with mannequin heads), and the jurors are often asked to rate which sound is preferred (paired comparison) or to rate a specific attribute (i.e. 'how whiny-sounding, on a scale of 1 to 7, is this sound?"). Thus, in research, making the recordinga s accurate as possible and therefore, as realistic as possible, is a requirement. True...your approach could be 'more accurate' then existing binaural, but to my knowledge, I have not seen your transducers supplanting the binaural mannequins. Note: That is NOT a slam on the fidelity of your system. Serisouly, it si not. It is however a question as to the accuracy of your system vis-a-vis your claim that it is technically superior to binaural and thus, yields better spatial resolution than does the currently held de-facto standard, the binaural mannequin head.
I don't mean to call into question your research, but all that I know has come from hands-on use of various mannequin types (I have used them all, save for one), and from performing both product sound tests as well as tests against each other using controlled lab conditions as well as live recordings. I feel I have a pretty good handle on the different issues and how each manufacturer addresses them in their mannequins. I have been lucky enough to work in acoustic analysis, sound quality, and noise and vibration as well as perceptual studies over the past 20 years. I know and work / have worked with a great many psychoacousticians and have had the benefit of having one-on-one conversations about much of what you have called into question. So, from that angle, I am have been very lucky indeed.
I know that there is a lot of debate about ear shape etc. I addressed this early on in one of my posts. I do not know what the correct ear shape is, save for one's own (and indeed, this is why there is an ISO working group that is still
trying to suss this issue out). I also concede there is no 'universal' ear and thus no universal HRTF. Indeed, in a 'perfect' world every person out there interested in binaural would have his / her own in-ear variant, complete with positioning gages (because changes in axial position would result in signifcant shifts in local maxima / minima in the left and right spectra, and potentially, temporal aspects as well).
As far as how the heads are equalized (I think
this is what you were referring to), there are three common approaches used:
* Free-field (designed to to accurately measure sources in a free-field, normal to the ear (90 degrees). This is more important for research and not all that related to music.
* Diffuse-field (designed to produce a frequency response in a diffuse field. Unlike free-field, this taked into account sind incident upon the ear from all angles)
* Independent of Direction (Taking into account only the effects of the ears and canals - not taking into account the diffraction caused by the pinna and the head proper)
These are the three main equalization schemes that are used on pretty much all commercially-available binaural mannequins (and they produce arguably very different results).
Basically, what is required to accurate localize sound are a few things, but they can be lumped into two categories
1) those that are a function of direction (diffraction, pinnae shape etc) and the presence of the person in the sound field, and...
2) those that are independent of the person's position (such as the auricle cavity and the ear canal)
As far as development, I believe that it was around 1980 or shortly after that that saw the first binaural mannequin that could be calibrated (I'm pretty sure this was the work of Dr. Klaus Genuit of Head Acoustics) and then further refined in the late 1980's.
Anyway, it seems to me that your approach has merit based on meeting criterion #1; the person's head and ears are present in the space - but this is also valid for in-ear binaural microphone methods (those worn by a person inside of the ear canals). However, so too do binaural mannequin microphones meet this crtierion - your approach, the in-ear approach, and the mannequin microphone all
satisfy the first criterion. As far as criterion #2 is concerned, that is arguably an average (as I understand it) because of the huge range of Human diversity)
So where does that leave things?
Well, I have to concede one point: I have not seen the data from your apparatus, so I can't say how yours might vary / compare when tested against a KEMAR, Aachen, B&K, Neumann, or a Cortex. However, one thing of which I am sure is that the differences could be described by the impulse response. I do think the recordings you presented sound very good though.
Another way to look at it might be via a controlled juried listening test. That is, I can see using your rig to record something (a variety of things by the end) at the same time and in (almost) the same location as a conventional mannequin head; if the space is large enough, I can see having someone wear in-ear mics positioned at the same approximate location as the mannequin / your device...though shadoing could prove problematic.
Ideally though, the best way would be to have a controlled test with your transducer placed and the source signal excited, then another transducer placed where yours was, then removed, and the next, then the next, and so on. As far as the listening test (juried test), it would pretty much have to be one wherein the jurors are asked to rate the spatial accuracy of the recordings...but how does one do this? One way would be to test againt human percpetion in an anechoic chamber
I mentioned this in another post, but there is a great little article in the March 2010 edition of Physics Today in which Hartmann (Michigan State University) and his colleagues conducted an experiment using test subjects in an anechoic chamber. Anyway, it's a great little article and if your transducer really does what you claim, then the playback should be in lock-step with the jurors' abilities to localize the sounds in the test chamber.
Again, I'm a data-driven kind of guy...and while I am not saying what you have done is not groundbreaking, I think that if you are going to make those claims, you should be open to having your transducer tested in a very controlled and structured manner with the participation of some willing university / reasearch organization. If you have accomplished all that you claim you have it would all be borne out by the experiments / listening studies. Frankly, if your approach is indeed proven
to be technically more accurate than binaural, then I would imagine that your transducer could supplant every
binaural mannequin head in every research institute around the globe (not to mention product development company (do you have any idea how many mannequin heads are used at companies that make cars, lawn mowers, washing machines, computer peripherals, guitars, pianos...the list is endless). As such, should your claim be proven
, you're knocking on the door to a huge market in perceptual research as well as product development.
Having worked in industry for many years, I can tell you it's not at all uncommon for new test equipment to be purchased for several hundred thousand dollars (budgets are typically much bigger than what you find in many if not most recording studios), but the wares that garner such purchase orders are those that are the accepted standards / state-of-the-art. Get your product into that
category, and...well...I think you get the point.