SO, this one is a little more fun than usual. Its a single-channel EEG headset called the NeuroSky ‘Mindset’ and its an interesting concept. The idea is to use the ‘mindset’ as a BCI (brain-computer-interface) for video games. Mattel has taken off with the idea, and according to some rating system, was given toy of the year for it. As a neuroscience student, i am happy to see people interested in this area, but i have to put out a warning, this thing has its limitations -and they are significant.
There are apparently several versions of the ‘mindset’ floating around and not all of them are compatable with the latest software. My headset was given to me by the Ramachandran Lab and the Pineda Lab to develop off of. I guess its an older version, so it doesnt work so well with the newer drivers. So, Im limited to 1Hz resolution of already fast-fourier transformed data (power levels for 8 different ranges: delta, theta, high alpha, low alpha, high beta, low beta, high gamma, low gamma). Not so useful for research that generally uses raw data at 100-120Hz. Although it is possible to get the raw data at a decent frequency, I am still not to keen on the fact that its a single electrode and theres no real way to measure impedence. The labs have their own ideas though, we’ll see how they turn out.
Theres apparently a few pre-processing steps that abstract the data coming from the device from anything natural. Its best explained below.
Taken from the NeuroSky website:
“Typically, power spectrum band powers would be reported in units such as Volts-squared per Hz (V^2/Hz), but since our values have undergone a number of complicated transforms and rescale operations from the original voltage measurements, there is no longer a simple linear correlation to units of Volts. Hence, we do not try to label them with any conventional unit. You can think of them as ASIC_EEG_POWER units, if you must.
The reason we say they are only meaningful compared to each other and to themselves is primarily due to the fact they have their own units as described above. It would not necessarily be meaningful nor correct to directly compare them to, say, values output by another EEG system. In their currently output form, they are useful as an indication of whether each particular band is increasing or decreasing over time, and how strong each band is relative to the other bands.”
The bottom line is that while the headset may be cool and fun, its not a research tool. There are other headsets, particularly the emotiv epoch system, that are closer to a potential research apparatus. I really like the emotiv system. It has more electrodes, gyroscopes, has algorithms for decoding facial emotion, and seems to be a more rigorously tested piece of hardware as all of the specs are available - and they seem to translate into real units. Also it boasts Allen Snyder, a famous neuroscientist, as a co-founder. In addition, the research edition is compatible with the standard EEGlab matlab toolkit. Total cost for the research edition is $750 (the consumer version is only $299, but doesn’t include the cool software that comes with the research ed). The neurosky mindset is consumer-priced at $199. Ill probably use some of the money that i earned developing for this project to either buy an emotiv or build my own ‘better than the rest’ version.
However. This marks the very first neuro-feedback application I have ever written. I plan to design and build my own wireless eeg headset with active dry electrodes and get the whole 10-20 system in. Once this is built, I can come back to my software and make it better.
My software currently does:
connects to the port where the bluetooth is forwareded from
does a handshake
checks for errors
if the data is good, parses the packets
outputs each received data point into a csv file
bar charts the power spectrum so you can see the values change
allows you to scale the various frequencies to the window
sweet