AES Headphone Technology Conference: Weird Science
Most of the papers presented at the conference were technically impressive and intellectually challenging, but, for the most part, not unexpected. There were a few, however, that were completely unexpected on my part. Here's three surprising and interesting papers.
A Low-Power Programmable Completely-In-The-Canal (CIC) Hearing Aid for Auditory Neuroscience
This paper documents the development of a miniature programmable completely-in-the-canal hearing aid ("active earplug") aimed as a tool for auditory neuroscience research. The main motivation of this project is to provide researchers with the ability to chronically change chosen aspects of the auditory experience in animals or human subjects. The active earplug is designed around a compact system-on-a-chip package (Belasigna 300, ON Semiconductors, Phoenix, USA) comprising digital converters, preamplifiers and a DSP processor. This chip interfaces with miniature receiver and microphone (Knowles, Itasca, IL, USA) with minimal supporting circuitry. The active earplug is designed to be custom-fitted in each subject's ear canal using a silicone mold. Arbitrary signal processing algorithms can be implemented on the DSP, therefore modifications of the acoustic inputs of the ear with an earplug can be chronically tested. This active earplug will enable researchers to study the effects of chronic expose to arbitrary signal degradations or augmentations, including hearing aid algorithms.
The presenter was unabashedly honest at the beginning of this presentation...this was a cry for help.
The paper was presented by a group of neuroscientists who are studying plasticity in the auditory areas of the brain. Evidently, when children experience recurring otitis media (ear infections), the acoustic filtering of the infection will cause the brains auditory areas to grow in unnatural ways. Once treated and physically cured, children may continue to have hearing problems due to the improper development of the auditory system in the brain.
What is currently not understood is what types of changes to the incoming audio signal cause what types of malformations in the brain. The researchers in this project have been developing a device that can be completely inserted into the ear canal of a chinchilla (the animal they use in these studies) that is able to modify incoming sounds in a controlled manner so they can carefully observe brain development with specific errors to the incoming signal.
What they found, unfortunately, was that at lower frequencies (under about 1kHz) the plug was not able to attenuate enough of the acoustically transmitted signal for them to effectively modify it using the device's electronic circuits. Their current thought is that some kind of noise canceling might be necessary to solve the problem.
One of the major challenges they face is that chinchilla ears are very small and the animals will scratch at their ears if anything protrudes, so the entire device, including the added NC electronics and microphones needs to be highly miniaturized. I hope they found a helping hand.
One important take-away from this presentation, which was not discussed in the paper, is that while brain plasticity is much higher in youngsters, it does exist with adults. As older folks begin to loose their hearing, the brain will try to accomodate; if hearing loss is experienced over long periods of time, the accommodation will become permanent. The point is, if you are beginning to experience hearing loss, it is much better to get hearing aids early on so that your brain doesn't rewire itself, which will make successful use of hearing aids more difficult later.
A novel prototype of an individualized electronic earpiece providing acoustic transparency, i.e., a sound impression that is perceptually equivalent to the open ear, is evaluated in terms of its listening quality. However, methods allowing for testing such advanced hearing devices in a way that is comprehensive, subjective blind, realistic and easy to use for the subject, are not readily available. We therefore present an according modular framework and use it to evaluate the new prototype operated as a personal assistive listening device in combination with advanced signal enhancement schemes. The introduced evaluation method is directly applicable for testing all hearing devices affecting a listener's live sound perception.
The main thrust of this paper was a new method for testing a hearing aid-like devices using a dummy head with remote listeners evaluating the sounds through headphones. But the thing that interested me most was the device itself and it's potential capabilities.
The device in question looks rather like a custom IEM. The bore of the device is open to the outside world allowing outside sounds to rather naturally be heard. But within the bore are are two balanced armature receivers and two microphones; one of each near the entry, and one of each at the ear drum end. There is also a microphone located in the concha part of the device. As one might expect, there is also a whole lot of DSP going on.
This device comes from the hearing aid world and it's main intention is to improve people's hearing acuity. But with the current trend in the headphone world for augmented reality capabilities, desired headphone and hearing aid performance requirements are beginning to converge. Augmented reality headphones will have a strong need for hear-through of the real acoustic environment and modifications thereofwhich, of course, is essentially what hearing aids do.
One of the very desirable capabilities for hearing aids or assisted listening devices is selective noise canceling. Trying to have a conversation with friends while seated at a table in a noisy, crowded restaurant can be very very challenging for someone with hearing impairments.
Two of the enhanced DSP algorithms evaluated in this paper are labeled as "interaural magnification and binaural coherence based noise reduction." The idea here is to determine how binaurally coherent a sound ismeaning how much a sound appears to be coming from a distinct position outside the headand to enhance signals that are strongly positional, and to likewise decrease noise that is more spatially diffuse. What that means for someone listening to dinner conversation is that the diffuse background noise of the restaurant will be reduced, and the positionally distinct voices of table-mates will be enhanced. A real boon for the hearing impaired.
"But geez,", you might say, "with all this fancy signal processing won't you end up with a pretty artificial sounding experience?"
The subject of this paper is a listening test to determine just that: How good was the quality of the listening experience? Researchers found that the new device is not only significantly better than a current, high-quality, behind-the-ear hearing aid running these advanced algorithms, it proved also to be moderately better than the open ear canals of the dummy head with NC algorithms applied after the fact.
During a break after the paper was given, I asked the presenter if these devices could be used to essentially give people "bionic hearing". He said indeed there is, and similar research in assisted listening devices for fire/rescue and military applications is ongoing.
You can think of this device as a sort of telescope/microscope for the ears!
When ANC (Active Noise Cancellation) specialists are asked about the purpose or benefit of ANC they generally answer "noise damping, isolation, reduction etc". Receiving the counter argument why not use earmuffs, potentially in combination with earphones, make one realize that the principle benefit of ANC is much more a matter of wearing comfort. Noise isolation as good as in earmuffs for industrial applications is offered by a small, lightweight comfortable headphone. This is realized by augmenting the passive noise isolation of a headphone by active noise reduction. Here, typically a closed headphone is used. Extrapolating the primary benefit of ANC headphones, leads to the vision that ANC headphones in the future should mainly gain in wearing comfort. This paper deals with a first technological approach towards this target. A semi-open headphone with its typical wearing comfort advantages providing the broadband noise isolation of closed earmuffs.
I really enjoyed this paper because it did a great job of bringing out all the nuanced difficulties in designing an active noise canceling (ANC) headphone. ANC headphones are typically designed to provide isolation below 1kHz through active noice canceling, and by passive acoustic seal above 1kHz.
ANC circuits have a difficult time working above 1kHz for two reasons: First, outside noise sources coming from differing angles will have different path-length differences between the outside microphone and the ear canal entrance, making feed forward designs more tricky since there isn't a fixed delay between the outside mic and the ear canal entrance. Second, the stability margin of the feedback ANC filters become unstable as frequency increases so the gain of the circuit must be reduced, which in turn reduces the ANC effectiveness.
The other problem addressed in this paper is that sealed headphone have some inherent comfort issues: pressure build-up in the headphones can lead to a disorienting feeling; the occlusion effect increases footfall and cable born noise; and a sealed headphone will feel warmer and is less able to vent moisture than an open headphone.
In a lovely example of out-of-the-box thinking, Sennheiser engineers explore the idea of using a semi-open headphone to solve the above problems. The headphone (see image above) has an opening in the capsule housing that allows external sounds into the headphone. A microphone is placed at the entry hole. Sound coming through the entry hole goes around a "delay plate" in the rear cavity before reaching the ear.
Now, because outside sounds enter through the hole rather than vibrate through the body of the headphones, sounds from various sources always take the same amount of time to go past the mic and to the ear. This fixed path length allows engineers to set one fixed time delay for the feedforward noise canceling circuits making them more effective, increasing the cut-off frequency of ANC to rise from 1kHz to 7kHz. And because of the open design, the occlusion effect is significantly reduced, and heat and moisture venting is significantly more effective.
The end result is about the same amount of noise canceling as available in a high-quality ANC headphone, but wearing comfort is improved.