If you’ve ever seen an experienced radio operator pull a signal out of the noise, or talked to someone in a crowded noisy restaurant, you know the human brain is excellent at focusing on a particular sound. This is sometimes called the cocktail party effect and if you wear a hearing aid, this doesn’t work as well because the device amplifies everything the same. A German company, Fraunhofer, aims to change that. They’ve demonstrated a hearing aid that uses EEG sensors to determine what you are trying to hear. Then it uses that information to configure beamforming microphone arrays to focus in on the sound you want to hear.
In addition to electronically focusing sound, the device stimulates your brain using transcranial electrostimulation. A low-level electrical signal tied to the audio input directly stimulates the auditory cortex of your brain and reportedly improves intelligibility.
The company isn’t producing the hearing aids yet but is working with the University of Oldenburg to bring the devices to their full potential. Although they are working on making the device more like a conventional hearing aid, it is difficult to imagine that you would not have to wear something over your head. Perhaps the EEG part could go in a sock cap.
The company claims there may be other uses for this technology in the medical field or safety-critical work situations. We aren’t sure what that means. Perhaps to know that you are actually listening to something or detecting that you are dozing?
We wondered if such a device might work out for seizure detection. If you want to do your own experiments, don’t forget about OpenHardwareExG.
Photo from Fraunhaufer website; credit: University of Siegen, Tim zum Hoff.
I’m VERY skeptical of those claims. Detecting the direction someone is trying to focus on is probably a couple orders of magnitude harder than what people have been doing with EEGs around, and then they want to mix that with tDCS which will further degrade the signal?
Sounds more to me someone was digging for some grants to play with.
I think it might be trying to detect the audio signal in brain signals, correlate it with sources and amplify the one that causes most response in the brain. Essentially detecting how the brain filters the signal and adjusting the prefilter to match. But I have no idea if it actually works.
Yes, they’ve been working backwards up the subvocal tree from throat mics, so possibly each level they can walk that back gives insights or recognition patterns that can start the ball rolling for further up the nerve tree.
I totally agree with your skepticism.
The numbers of neurons that the apparatus is trying to analyse run into the billions. Writers of these articles need to look at better science..for example current studies use Squids [superconducting quantum interference device magnetometer]. These devices can look at columns of neurons and can stimulate as well.
Squids are large, need to be very cold and only found in research institutes.
Yet here is a company who can do better with a cheaper product. Smells like the Scientologists e-meter – a simple ohmeter measuring galvanic skin responses!
And from my experience in anesthesia over 30 plus years, many companies have made devices to try and tell if the patient be asleep or not…nothing that can do that yet.
Don’t waste your money or time on vapour/neuralware.
But is it so much “finding what direction someone is thinking” bht ” training someone to indicate something”?
I was under the impression that with artificial limbs that are motorized, they don’t try to map directly to old use of muscles, but train the user to generate a signal that can be used. Tbus there are lots ofsignals available, pick one and train.
“If you’ve ever seen an experienced radio operator pull a signal out of the noise, or talked to someone in a crowded noisy restaurant, you know the human brain is excellent at focusing on a particular sound. ”
Except when someone else is flapping their gums right next to my other ear.
Interesting, thanks for sharing. Great idea with the feedback design concept. Optimization of each specific task I’m thinking is interesting in itself to work on.
Really is about time a system came out that doesn’t require an implant to tap deeper down the audio train.. and I suppose a minimally invasive passive sympathetic resonator is the most I’d consider to prevent morbid operations if the surgeons are a risk.
Why do that though, when the science and technology is out more main stream to be able to beam form voice into the head in more than one ways and means… why not beam form into the cell cluster specific to the auditory system or maybe even deeper down the signal path for those with nerve damage?
Wondering how specific the signals are to individuals or if there will be a required training step process to calibrate the system for users specifically… I guess with some quantitative validation requirements too.
About time… even if propaganda. :-|)
How about using a tilt sensor to indicate the direction of the sound you want to hear. The user just tilts their head in the direction that they want to emphasize. It’s a natural reaction when you lean in to hear someone better. The speed of the tilt, direction and amount of tilt could be sensed to vary the focus of the beam forming microphone array. No fancy EEG processing necessary.
If you tilt/rotate your head, what would beamforming be useful for in this case? I thought beamforming was useful to direct waves around a stationary point, similar to SAR
Fraunhofer? They are those with the MP3 patents, aren’t they?
I regret every cent of my tax money which went in their general direction (otherwise I do pay and I do support paying taxes).
The MP3 patents didn’t actually seem to stop much. Some big companies might have invented some more proprietary nuisance formats because of them, but the FOSS community mostly reverse engineered those too.
Is there any cases of it actually being a real problem?
Yes, stalling the field for at least 5 years. Reduction in technological progress is the real problem with patents, not a problem in 1910 but in 2020 a patent can bottleneck progress for up to 20 years unless there are alternative approaches.
Cough. So … you hate mp3, I get that. But it has been (and in some countries still is) a HUGE success. If Seitzer and Brandebburg hadn’t developed it – they actually invested WORK, you know, that horrible exercise nobody today wants to pay any money for – mp3 would not exist.
As for Fraunhofer: Just because they held the patents, that’s not bad. Patents can be and most often are a GOOD THING ™, if they are being used the way the idea of patents had been invented itself. Which, in the case of mp3, has been the case: An mp3 license was and is EXTREMELY cheap.
BTW: Fraunhofer’s own mp3 player was a complete pipe shredder. It was mainly thanks to Win AMP that mp3 made it. Again, without mp3 having existed, a lot of modern music/audio compression would not be as advanced as it is.
That is not the fault, but thanks to Fraunhofer. They are one of the few “good guys” out there, inventing, developing and researching in so many areas, including fundamental research that is required and used by many, many others, that every single penny invested into their work is well invested. Including when their projects don’t turn out successful, because SOMEONE has to do the work. They do.
Fraunhofer is not a company, they are a society ;-) If someone (in Germany) can pull off something like that, it’s them. Nice to see a contribution of them on hackaday.
Trying to read where focusing attention could be interesting, but may take a bunch of work.
In the meantime – how about offering some lower-tech steering alternatives.
* Eye trackers (glasses/etc. which observe where I am looking, and focus in that direction).
* Track a target – give me a target designator (e.g. an object I can pass to whoever is speaking).
(In restaurants one would usually want to hear somebody nearby at the table.)
Unless this is going to control something like volume or frequency, it surely relies on two hearing aids, since two ears, with a level of isolation between them (the head) provides directionality.
Do they really know how a signal can be picked out of the noise (white or interference)? Because while there have been projects to add a “stereo” effect for listening to signals on the radio, generalky it’s all mono, so something else is happening in the mind.