Greetings, Mostly through serendipity, I have had the pleasure and privilege of great teachers. I studied recording arts under Andy Seagle (andyseagle.com) who recorded Paul McCartney, Hall & Oats, and numerous others. My doc committee included Bill Yost, who is widely known among the spatial hearing folks. And, of course, I've learned a lot about Ambisonics from people on this list as well as a plethora of technical articles.
I recently sent an email to Bill with the following question/scenario. I thought others might wish to give this thought, too, as it gets into HRTFs. There have been a lot of studies regarding localization in the transverse (horizontal) plane. We also know from experiments how well (or poorly) we can localize sound in the frontal and sagittal planes. By simply tilting someone back 90 degrees, his/her ears shift to another plane. This is different from shifting the loudspeaker arrangement to another plane because the semicircular canals are now in a different orientation. If a circular speaker array was setup in the coronal plane and the person was lying down, then his/her ears would be oriented in such a way that the speakers now circle the head in the same fashion as they would in the horizontal plane when the person is seated or standing. It's a "static" vestibular change, and gravity acting on the semicircular canals (and body) lets us know which way is up. But do we have the same ability to localize when the body is positioned in different orientations, even when the sources "follow" the orientation (as is the case in the above example)? How about localization in low-g environments (e.g. space docking)? The question came to me while camping. I seem able to pinpoint sounds quite well in the (normal) horizontal plane despite a skewed HRTF while lying down (and somewhat above ground). On another (but related) topic, I have downloaded the HRTF data from the Listen Project, and have been sorting the participant's morphological features. I have this in an Excel spreadsheet, and am converting this to an Access database. Using the data, one can pick an "appropriate" HRTF starting with gross anatomical features (such as headsize) and whittle it down to minute features (such as concha depth or angle). I find HRTF discussions interesting, but still argue that headphones and whole-body transfer functions make a difference, too. Insert phones destroy canal resonance, whereas an earcup with active drivers may have a large "equivalent" volume, thus minimizing external meatus/earcup interaction (a mix and match of resonances). Because of this, there can be no ideal HRTF, even when it matches the listener. While listening to HRTF demos, the notion of auditory streaming and auditory scenes came to mind. Some sounds were externalized, but other sounds of varying frequencies, while emanating from the same sound source, appeared in my head. The end result was that the externalized sounds provided a convincing (or at least fun) illusion, but problems do persist. A stringent evaluation of HRTF / binaural listening via headphones would require breaking the sounds into bands and seeing if a sound's constituent components remain outside of the head. When doing so, a brick-wall filter wouldn't be necessary, but a filter that maintains phase coherency would be recommended. The demo I refer to was that of a helicopter flying overhead. Though I haven't done this (yet), it would be interesting to use FFT filtering to isolate the turbine whine (a high-pitched sound) from the chopper's blades. The high-pitched sound appeared to be in my head, whereas the helicopter as a whole seemed externalized. Again, an individualized HRTF and different phones may yield different results. Side note: Be careful using FFT filtering--it can yield some peculiar artifacts. I am hoping to use headtracking in conjunction with VVMic to model different hearing aid and cochlear implant mics in space. This offers the advantage of presenting real-world listening environments via live recordings to study/demonstrate differences in mic polar patterns (at least first-order patterns) and processing without the need for a surround loudspeaker system. In fact, it's ideal for CI simulations because an actual CI user never gets a pressure at the eardrum that then travels along the basilar membrane, ultimately converted to nerve impulses. With VVMic and HRTF data, I should be able to provide simulations of mics located on a listener's head and then direct the output to one or both ears. This does not represent spatial listening, but it does represent electric (CI) hearing in space. Putting a normal-hearing listener in a surround sound environment with mock processors and real mics doesn't work because you can't isolate the outside (surround) sound from the intended simulation, even with EAR foam plugs and audiometric insert phones. VVMic and live recordings via Ambisonics is a solution to creating "electric" listening in the real world. Again, I'm referring solely to CI simulations. With the advent of electric-acoustic stimulation (EAS), more than one mic is used per ear: One for the CI and a second for the HA. Combinations of polar patterns can be created. Respective frequency responses and processing can be sent to one or two ears (diotic and dichotic situations). One caveat for using vocoding to mimic CIs is that the acoustic simulation (and therefore stimulation) still necessitates a traveling wave along the normal-hearing listener's basilar membrane. The time it takes to establish a wave peak is not instantaneous (though compressional waves in the the inner ear are virtually instantaneous), and I believe a time-domain component to inner ear (mechanical) action can't easily be excluded when using "acoustic" simulation of CIs. I suppose I could look at data from BAERs and the Greenwood approximation to account for the time-frequency interaction. Just some thinking... and ideas to share with others interested in hearing impairments. By the way, Teemko, if you're reading this, just wanted to let you know that Bill Yost said he'd read your thesis over the weekend. I notice that Bill and Larry Revit are in your references list. Larry isn't a fan of Ambisonics--said to me in a phone communication that it sounds "tinny". I suppose it does if one were to listen through laptop speakers or from poor source material. Not sure what his source was. -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://mail.music.vt.edu/mailman/private/sursound/attachments/20121103/837528f1/attachment.html> _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
