Inexpensive EEG Devices?

Rustcycle akss: "To extend prior music generation experimentation, I'm interested in creating music via genetic algorithms using neurofeedback to assign fitness values. Does anyone have a recommendation for EEG systems that are affordable outside research institutions? What's the best system under $2k? Ideally I'd want a multi-sensor system so I could do sonification experiments to 'hear' correlated data from different regions, but I'd settle for a one or two sensor system for initial experimentation — so long as there are drivers for Mac / Linux. How safe / unsafe is the OpenEEG route?"

Comments

Homebrew

[DrLang21]

Under $2000? I would be surprised if there was a system that cheap. I think your best bet is to build you own differential amplifiers with a couple of opamps.

Re:Homebrew

[smellsofbikes]

Instrumentation amplifiers. They're cheap and astounding. CMRR's of 80 or 100. Take a look at the Analog Designs AD620, for instance. It's superb (even if they are our competitors.) I've used it for making an EKG based on an old Scientific American Amateur Scientist article, and here's a slashdot thread about another AD620-based EKG.

This system is dirt cheap

[Centurix]

1. Buy 2 kilos of sugar and a hand-held Casio keyboard.
2. Eat sugar
3. Compose like your hands are on fire.

You need enough sugar floating around so that if you ever take a PET Scan, your brain would show up a sort of platinum white colour on the screen.

That depends

[nizo]

How safe / unsafe is the OpenEEG route?

That depends, how comfortable are you drilling holes into your own skull?

OpenEEG

[Anonymous Coward]

My senior design team from College actually went ahead and tried to build the machine featured on OpenEEG. In my experience, we were not capable of getting a signal with a high enough resolution of detecting anything other than "ACTIVE" or "NOT-SO-ACTIVE". But the circuitry provided does ..work... after a bit of tinkering. The resistor values are very important. Try to be as exact as possible. Also, shield everything you can from inteference. You'd be surprised how much interference the power lines in your house emit. It will be about around 60Hz, and periodic, if you are picking it up. The DRL portion of the circuit helps to reduce it.

As for analyzing the data it produces, that also becomes difficult. "ACtivity" on an EEG signal could be as small as a uV. Sample it as fast as you can. We used a PIC processor to sample.

Also, muscle signals can drown out the electrodes, try not to move.

A Friend of Mine

[TubeSteak]

A friend of mine bought a EEg for neurofeedback... but the sucker didn't come with any drivers, disks or manual.

And he can't find anything on the internet that is useful to get it working.

So, be careful what you buy, because you might just get a hunk of hardware, but no software to run it.. if you're going the cheap route that is.

Re:Maybe back issues of electronics magazines?

[danlyke]

Steve Ciarcia did a system back in the '80s in Byte magazine that I remember primarily because I was reading through it and thought "why's he using batteries and doing so much work to isolate this side of the board from the other?", and then realized the possible problems that could stem from power supply failure or getting RS-232 voltages on to the probe side of the board. But whenever I go back to those old circuits I'm shocked by the complexity of the designs, mostly because we now have a much better array of integrated circuits and complete devices.

I've looked at the OpenEEG circuit a bit, and while my hardware experience is limited to building things like stepper motor drivers I've considered building it and hooking it up to myself. If you have trouble with the analog side of the board you might try looking around in the old-school guitarists in your area; I've found a number of guys who played back in the '70s who got really good at analog electronics and noise shielding because they got into building their own pre-amps.

And as others in this thread and elsewhere have said, once you build the hardware you've done the easy parts. You're looking for fairly small low frequency signals in a world full of very big low frequency signals. So put a value on your time and know that a lot of what you'll be doing is signal processing on a bigger CPU, not just using a microcontroller to feed small measurements to that big CPU.

DON"T TRY THIS AT HOME

[MountainLogic]

Seriously, the subject MUST have full galvanic isolation from the power lines. The reason for this is that in every day life your skin provides a surprising amount of resistance and thus protection from shock. What is safe, 48 VDC, can kill an electrically compromised subject. When you put electrodes on the skin you create a low impedance path into the body. The typical solution for this is to use optical isolation. Check out Dallas/Maxim Semi and others for off the shelf solutions. DO NOT hook someone up to an oscilloscope - one leak in a transformer and they are dead. The other solution is to use batteries, but you have to be careful to limit the available voltage and current from a battery source.