Awesome new (relatively inexpensive) Brain-Computer Interface | Hacker News Reader

Awesome new (relatively inexpensive) Brain-Computer Interface

emotivepoc.com

57 pointsspeek28 comments

Please note that the developer headset and the EPOC consumer headsets are a separate line of hardware products and are not interchangeable. The developer headset supports development whereas the consumer headsets ($299) will only work with approved applications.

Or: "Please pay us an additional $200 for the chance to create anything that would create a demand for our product." No mention of the cost of "approval".

And you need to pay them anywhere from 750 to 7500$ bucks to develop an application for it...

Developer headset is 500 bucks, Research edition (developer headset with EEG access) is 750, or 650 if you bought a developer headset a while ago.

Agreed, although they're adjusting, slowly. I came across this earlier this year but they didn't even have a proper developer program at that time. One can get pick up used EEG equipment pretty cheap on eBay, however, and there are some open development tools from the scientific community that run on Linux, as I recall.

$300 is accessible to "hobbyist" researchers (hackers) - this is the main point of this story - and hackers will crack DRM.

I guess I should have read the fine print. I ordered the $300 device after they sent an email a couple months ago.

It supposedly has shipped and will arrive any moment - I didn't realize it would be cripple-ware

From the description, these are still the full headsets you get with the other editions, you just don't get the SDK. You might not have that much trouble hacking it. Once it's in your hands, you can start sniffing the USB traffic. They say it doesn't require a driver, so it's probably a USB HID device which is even better and easier to work with.

You might still get pretty far even without the SDK. Once it arrives, you could post here on Hackernews for some help. I have a bit of experience reverse-engineering USB devices, but I'm sure there are others here that could help even more.

Looks like you can achieve near 100% accuracy with one caveat - the electrodes have to be inside your skull, not on top of it: http://hardware.slashdot.org/story/09/12/23/1511251/Typing-W...

>>...an experiment involving two patients with epilepsy. Both patients were already being monitored for seizure activity using electrocorticography (ECoG), in which a sheet of electrodes is laid directly on the surface of the brain. This procedure requires a craniotomy, a surgical incision into the skull. Dr. Shih and colleagues hypothesized that feedback from electrodes placed directly on the brain would be much more specific than data collected with EEG (electroencephalography) alone, in which electrodes are placed on the scalp. Most studies of mind-machine interaction have occurred with EEG. "There is a big difference in the quality of information you get from ECoG compared to EEG. The scalp and bony skull diffuses and distorts the signal, rather like how the Earth's atmosphere blurs the light from stars," says Dr. Shih. "That's why progress to date on developing these kinds of mind interfaces has been slow."<<

I don't know, it seems an awful lot like listening to a conference hall for one user's voice commands in an unknown language. Brain-wave activity is largely unknown, not understood, and noisy as hell. There isn't even any concrete evidence that distinct portions of the brain are singularly or wholly responsible for specific tasks. Anything we do related to the brain in this period of human development is not much more than witch doctoring.

I'm only theorizing here, but I don't think its trying to look at a specific part of the brain, then determine what you're thinking. It seems to me like there are common actions people can perform that render similar results across everyone (for example it seems like smiling is one) if that is true then you simply need to train the computer what your particular brain looks like after performing this action. Once you have that, you simply map that action to your desired output. The concept is simple (though i'm sure the implementation is wildly complex)

The concept is simple (though i'm sure the implementation is wildly complex)

Those sound like famous last words :)

it seems like smiling is one

Smiling is perhaps a bad example, since it contracts a muscle near the sensors. When someone is smiling, it's probably hard to detect anything else.

  > Smiling is perhaps a bad example, since it contracts a 
  > muscle near the sensors.

This is no doubt intentional.

This is why you would likely have to "train" it.

edit:

here is a relevant link: http://www.youtube.com/watch?v=40L3SGmcPDQ&feature=relat...

They begin training it at about 13 minutes in.

Having done work with a similar sort of recording (fMRI, which is actually quite a bit more spatially sensitive to brain activity) I, like the op, don't have a lot of faith in "training" unless someone has developed a fantastic model of the brain's functioning and never talked about it.

That being said, brain-computer interfaces are probably decently possible if they abuse brain plasticity to overload our already highly developed IO systems: our senses and physical actions.

It's exciting to think where this could take us in 10 years.

Remember the first mp3 players? http://upload.wikimedia.org/wikipedia/en/0/0b/Rio_pmp300.jpg

Now we have stuff like this: http://i50.tinypic.com/30cn1va.jpg

Computer assisted telepathy, here we come.

you are comparing a mp3 player with a tiny computer, would make much more sense to compare it to a large computer

i'm dubious these things can accurately interpret anything more than face muscle tensions, eye-movement if that. which are common noise sources in EEG recordings

http://en.wikipedia.org/wiki/Electroencephalography#Artifact...

anyone have some anecdotes otherwise? the demos on youtube are not really satisfying.

You can eliminate noise/artifact like that by subtracting it from a reference point.

but they don't seem to be doing that since "expressive" is one of the detection aspects they mention.

I think that's just "eyebrows went up, smiling muscles contracted, blinked", etc. for the look-my-emoticon-changed marketing effect. They also claim there's a gyroscope in the device.

The more advanced EEG tech is quite real. It's a stretch to say that it's interacting with your "thoughts" ... making something like PacMan move forward/backward is a similar sensation to when you first figured out how to wiggle your nose (the mental exercise, not the muscle movement).

Looks nice, and I've seen the Marvin Minsky demo, but the fact that there is no Linux version means it would probably be useless to me. I'm not sure how practical these interfaces will turn out to be, other than for entertainment purposes.

The fact that there is no linux version is no impediment to determined hackers.

I've taken various classes in Neurofeedback interaction design. The biggest obstacle to getting work done was the frustration of hooking electrodes up to your own scalp. Also, circa 2005 the lowest barrier of entry for interactivity was through a Flash gateway in the BioExplorer IDE (http://www.cyberevolution.com) ... whereas now I think with Air you could send event messages to something like processing.js, flot, jQuery, HTML5 <audio/video>, sparklines, etc.

I'm an EEG technician, with a background in electronics and programming. I use BioExplorer for neurofeedback routinely.

I suspect that the emotiv system relies on conditioning (eg. if i do action "A" - the computer moves the pointer up on the screen), not real EEG pattern recognition.

For the record, BioEra (www.bioera.net) does have support OCZ NIA system - cheaper, and no developer road-blocks.

BTW Your 100% on the mark about reducing artifacts. There is no substitute for a good connection to the scalp.

Edit: www.bioera.net for details

@neurotech1 : Thanks for the BioEra link. Hadn't heard of it.

Have you run into any development tools that will allow you to send threshold triggers quickly enough to a web browser rendering engine?? (like BioExplorer does with their Flash gateway)

You could probably hack Javascript function to work with the BioEra XMLServer. Something vaguely similar to AJAX but I wouldn't consider it "standard" AJAX programming.

> There is no substitute for a good connection to the scalp.

Most people don't realize that large pieces of coral, which have been painted brown and attached to the skull by common wood screws, can make a child look like a deer.

Reply on news.ycombinator.com