What kind of things can you do with the more consumer oriented EEG headsets on the market? Here is a look at the lay of the land right now.
The typical EEG devices for research will give you access to the raw signal along with a suite of analytical tools (read here about the EEG signal and what it measures). For example g.Tec’s EEG toolbox has tools for data quality estimation, power spectrum and bandpower estimation, wavelet analysis, even event-related coherence, desynchronization (ERD) and synchronization (ERS) and more. Advanced Brain Technologies also has a LORETA package for source localization that comes with their hardware.
The more consumer oriented research headsets (defined largely by pricing) come with software that is less developed in terms of analytical capability, mostly just providing the raw signal and FFT. On the other hand the price points and/or greater ease of use opens up new possibilities for application development. Here’s a look at how they work right now from the type of raw signal you get, the associated softwares and SDKs.
|Emotiv EPOC (14 channels) and Insight (5 channels)||PureEEG: Access to pre-filtered raw data from the headset and FFT, online monthly subscription based software. An annual subscription based offline version is now coming. No formal SDK, must contact company||EDF||Windows, Mac, Android, iOS||$49 /mo for 50 recordings||Talk to company|
|Open BCI UltraCortex Multiple configs from 4 to 32 electrodes||OpenBCI GUI: Access the raw data with tools to view raw data and Power Spectrum and perform offline filtering. Developer tools and templates available for Java, Python and Node.js||EDF||Windows, Mac||Free with device||Tools and templates for development with Java, Python and Node.js|
|Neurosky Mindwave Mobile and Mobile Plus (single channel)||Neuroview/NeuroskyLab: Access to pre-filtered raw signal, FFT and processed metrics (attention and meditation). NeuroView provides CSV output, NeuroskyLab integrates with MATLAB where you can use other tools like EEGLAB.||CSV||Windows, Mac, And, iOS||500||Free, standardized toolsets for all platforms with raw signal access|
|Neurosky MindWave Mobile||MyndPlay Pro: third party app providing access to the raw signal, real time frequency bands (theta,alpha, beta, gamma)||CSV||Windows, Mac||121.74|
|MUSE (4 channels)||MuseIO and MuseLAB: Access to raw signal and FFT.||MUSE format||Windows, Mac, Linux||Free|
Emotiv EPOC is a 14 channel headset that uses gold electrodes with sponge contacts that are wet with saline, while Insight has 5 polymer based dry electrodes. Emotiv has discontinued their downloadable TestBench software for accessing the raw EEG signal from EPOC and moved to a subscription based model that supports both EPOC and Insight. Called PureEEG, this online software does much the same thing but requires Internet access to use it and comes with a monthly subscription, which is a totally different pricing model. For folks like us who use the device in remote locations with no Internet they are working on an offline mode with an annual subscription. PureEEG does basically the same thing as TestBench did though it actually has fewer features – for example, it does not come with the Push/Pull training and other apps that TestBench was bundled with. Rather this is meant purely for raw signal access giving you a view of the raw signal as well as the FFT. Files are saved in EDF format, which you can easily import elsewhere for analysis. The most developed EEG libraries are of course available now for MATLAB (EEG LAB) and various repositories on GitHub but there are now some libraries for R.
Some things to note – EPOC uses mastoid references and provides the raw signal after a bit of processing – they filter between 0.1 and 100 Hz and then have notch filters at 50 and 60 Hz so that’s what you get as an output. You can read more about the Emotiv EPOC here and an evaluation of its signal quality here.
The subscription based pricing model is based on the number of recordings that you do. You pay $49/month for 50 recordings, ~ $1/recording. If you do 250 recordings a month you pay $149, which is 60 cents per recording. You can also prepay for the year and get some pricing benefits – for 50 recordings a month with the annual prepaid, you pay 83 cents per recordings rather than $1. The big downside is that if you make mistakes in your recordings and need to scrap one and start over you end up paying for all that – on the other hand it can make you more diligent about your data collection.
In terms of developing new applications that work with the Emotiv devices, there is not yet an established SDK or application developer program. At present you have to ‘apply’ or basically talk to them to develop applications. Presumably this will change someday.
OpenBCI’s UltraCortex comes in various configurations from 4 to 32 electrodes and has traditional gold electrodes used with conducting paste so would therefore have good contact quality. While the device is priced higher than EPOC, the software is free to use. The raw signal they provide is less processed than the Emotiv – you get the raw signal and then have the flexibility to do the processing in the software GUI. For example, the GUI has options to apply your own filters to the data including band pass filters and notch filters at 50 and 60 Hz. The default is 1-50 Hz. In general, the Open and DIY nature of OpenBCI allows you to find ways to get at the signal at any stage of processing (even the internal signal) if you really want it. Like EPOC it is just designed to give you access to the signal and in terms of analytics it provides only the FFT. Also, here again files are saved in EDF format and can be analyzed anywhere. You can read more about OpenBCI’s Ultracortex here.
The OpenBCI has some SDK tools that are basically a set of templates and tools to develop applications for OpenBCI in Java, Python and Node.js that are available on their site and on github. This is still really an open community focused on DIY hardware so it’s not really driving towards a consumer application marketplace and is more a playground for hackers.
Neurosky is fundamentally different from Emotiv EPOC and OpenBCI: first it is a single channel system (with a mastoid reference and a ground) that uses a dry stainless steel electrode (that presses too hard on your forehead if we may say so) and provides a pre-filtered signal sampled at 512 Hz as output (compared to 128 and 125 Hz for Emotiv and OpenBCI respectively). Neurosky’s device Mindwave has been built for consumer applications, not research, and they have a well-established app marketplace and SDKs for iOS, Android, Mac and Windows. The developer tools are free and also give you access to the raw signal. Neurosky also has a standardized platform and revenue share structure for selling apps in their store.
That said they do have software for research that, like Emotiv’s PureEEG and the OpenBCI GUI, provides access to the raw signal as well as power spectrum. In addition they also provide you their own ‘eSense meters’ for attention and meditation though how these are computed and what they really tell you is not clear. There are two different versions – Neuroview, which does not provide an EDF output but rather simply gives you the signal as a CSV and NeuroSkyLab which integrates with MATLAB. Both cost $499 and only work with Windows.
There are also third party research tools compatible with Neurosky. MyndPlay for instance provides all of the same basic features as Neuroview and Neuroskylab but is only $121 and works on both Mac and PC. The MP tools also include other gaming and neurofeedback training and one other BCI applications — an audio / video media player function for observing your brain’s real-time reaction to media. Here it is essential to check headset compatibility before purchasing as it may not be compatible with the Neurosky MindWave Mobile-plus headset –their latest and most promoted headset! This is confusing because the same apps (almost all of them) ARE compatible with the MindWave Mobile (no plus) headset.
Muse, which calls itself the ‘brain sensing’ headband, has 4 dry electrodes, 2 silver electrodes on the forehead (anterior frontal) and 2 conductive silicone-rubber electrodes behind the ear (temporo-parietal). Maybe a step up from the Neurosky stainless steel in terms of signal quality but not by too much. They use electrodes on the forehead (FPz; CMS/DRL) for reference. The raw signals can be accessed through their developer tool kit with sampling rates set to either 220 Hz or 500 Hz and is low pass filtered at 100Hz. The 60 Hz notch filter which is turned on in the consumer application is switched to off for the research applications.
To get the raw signal for research however requires two separate softwares: MuseIO, which links to the device and MuseLab, which streams the data from MuseIO and makes it available. The software then allows you to record and save the raw signal but it is anything but easy and intuitive. The data is saved in a .muse format. You will need yet another application – MusePlayer, which will convert it to CSV or MATLAB format if you want to use the data with EEGLAB or any other toolbox. Basically, it’s all enormously complicated and has far too many steps to get what you should be able to get with a single click.