# How to build a high sensitivity EEG headset for continual monitoring?

After having a conversation with a couple of the more popular "consumer" level EEG makers such as Versus, I have found that their monitoring abilities are either limited or altogether absent. The devices utility, while cool from a 3rd party controlled training standpoint, is neutered when it comes to just getting raw data and plotting it out in a manner that allows one to make their own interpretations and calls.

I would like to build my own device that can monitor from the very low < 0.5 hz all the way up to at least 200 hz. My goal is to feed that into my own software in order to keep a running graph for analysis, etc.

How would one go about building an EEG that has this level of sensitivity?

There are a few commercial EEG makers that do allow monitoring of raw EEG data. For instance, I know the Muse headband and OpenBCI allow you to do this with little trouble. These both support sampling rates up to 250 Hz. (The OpenBCI in principle could have a higher sampling rate, but i don't know anyone who has done this.)

The Emotiv EPOC also has an option to get raw EEG data. You have to pay more for the research version of the SDK to get the data officially. However, someone has reverse engineered their protocol, so you can access the raw data that way. This has a sampling rate of up to 128 Hz.

That said, if you still want to build your own EEG, OpenBCI has documented all of their plans, so a good start would be to replicate their product.

There is also the older, but still useful, OpenEEG project, which also has open plans for an EEG device.

• That's really interesting, I am grateful to see that about the 250 Hz on the Muse. I'm curious; the Emotiv devices seem a bit more sophisticated, but the lower rate intrigues me. Any idea as to why they would have opted to go this low vs the higher Muse? I am thinking the other open options seem to be a better choice. Jan 1 '16 at 5:06
• With a 128Hz sampling rate, you can analyze signals up to around 45Hz (divide by three instead of two because antialiasing filters aren't perfect). That will get you all the way up to gamma, which is about as high a frequency as you can reliably record from a healthy individual on the scalp. I'm guessing that is why Emotive chooses this lower sampling rate. Some people try to record higher frequencies, particularly in epilepsy research, but it's such a tiny signal on the scalp that you would need an exceptionally noise-free environment and high quality amp to pick it up.
– K A
Jan 1 '16 at 6:06
• @serk1 this is very interesting, thanks for this input. I would be curious as to further expansion upon, therefore, the validity of the Muse unit. Is it simply offering a high sampling rate to look good, or is there value therein? It would essentially go up to around 83Hz according to the logic you explained there, so are you saying then that anything past 45Hz is simply going to be invalid? Jan 1 '16 at 18:53
• Neural activity past 45Hz isn't impossible to record on the scalp, it's just really difficult. Two factors contributing to that are 1) the skull acts as kind of lowpass filter and 2) high frequency neural activity is more localized than low frequency activity, and the skull also acts as a lowpass spatial filter. This makes signal to noise low for high frequency activity. The higher sr on the Muse could be helpful for better characterizing some of your noise sources (mostly muscle artifact) though. It's good to remember that most of the signal recorded from the scalp isn't coming from the brain
– K A
Jan 2 '16 at 5:24
• There's not much that I can add to serk1's excellent followup below. As serk1 said, high gamma is very hard to pick up with EEG, which is probably Emotiv's reason for this threshold. In addition, most work with EEG so far has been in the below 50 Hz range. Only recently have people started exploring picking up high gamma using EEG. Jan 2 '16 at 6:09