Alpha signals apparently random?


#1

Hi, guys.

The place I work for recently received a Muse, and I am trying to get some signals out of it, to use on some externals applications.
The first sessions I performed had the intention of learning about the device.
I have experience with OpenBCI and OpenEEG (built version by Olymex).

And the test I am using to “benchmark” the device is a simple Alpha testin:
Steps:
1- Open my eyes and talk for 5 seconds;
2- Close my eyes for 5 seconds;
3- Wait 10 seconds and repeat twice;

Using both the aforementioned devices I could get peaks on the alpha spectrum, and really identifiable ones. But using muse (MuseLab, running default command muse-io --device NAME), the waves for Absolute alpha for each channel seem to vary randomly, rather than following a pattern.

I already tried to clean my forehead with several products (the first idea was that it was caused by interference), and turning on and off the device. As well as trying another PC. Tried to perform the tests on a unplugged laptop (even though the DC offset from the wall plug won’t affect muse), and to no avail. Seems like the alpha Absolute is varying randomly.

I have already did some research, and found nothing like my question. =)

So, I may ask you:
1- Is this a problem with my device?
2- Is this the normal behavior of the waves?
3- Should I use conductive paste/gel when wearing Muse?
4- The approach to test the device responsiveness is wrong?

What else should I try?
Thank you in advance for your attention.
Best regards,
FernandoLL

p.s.: My main PC configuration;
AMD Phenom II X4 (965);
8gb RAM
Windows 7
MuseLab 1.6.3 -b17

Laptop
AMD E-350 (1.6GHz Dual Core)
4gb RAM
Windows 7
MuseLab 1.6.3 - b17


#2

An update on the situation: I spent the night trying to overcome this problem. Again, to no avail, but here are some of the changes I’ve made on protocol and software.
Updated the software to the latest version (as of now 3.4.1);
Now I’m trying to measure concentration and “mellow”, with the preset experimental waves;
Looking at the entire spectrum using the static line visualizer;

The lowest bands of the FFt still display an abnormal and irresponsive variation of the signals. With which I mean, when I close my eyes, and try to relax, nothing changes correctly, or the amplitudes go up and down as I blink. But not when I am relaxed.

Again, compared with results I had with OpenEEG, for a similar freq. range, Muse outputs are strange.


#3

Hi n4ndoz, I’m a fellow newcomer to the world of Muse but a well seasoned EEG user. Could the issue be ambient EMF in your test environment such as poor grounding, static charge from flooring, etc.?


#4

Hi, FranKiernan.
Sorry for taking too long to get back to the post.
Well, I don’t think my house would have strong EMF. By poor grounding you mean something on the PCs? About static charges, I still don’t think it is the case. Since I already tested at home and at the place I work. At both env’s the signal displays the same randomness.

I will try to use one of those “anti-static” wrist band’s to test your hypothesis. Since I’m not a “native” EEG user (I work mainly with molecular dynamics at the Uni) maybe some of the EEG tech guys at Uni could shed some light aswell.

Thank you a lot for the answer.

p.s.: What about the Muse’s tech support guys? They wonder around the forums?


#5

Hey n4ndoz,

I don’t think there’s a problem with your device.

A few things to consider:

The absolute and relative band powers are calculated for fairly short windows (we call them epochs), one every tenth of a second. From epoch to epoch it is quite normal to see a lot of variation. It’s only when you average epochs to smooth the data that you will start to see more gradual trends clearly. Can you try smoothing them?

The whole spectrum moving up and gradually back down when you blink or move your face is normal. These ocular and muscular artifacts can introduce broadband noise into the spectrum, meaning that power in all frequencies will increase.

In general, broadband noise will decrease over time as the impedance between your skin and the electrodes gradually decreases. There is a settling period with all EEG electrodes during which an electrochemical equilibrium is established. Perhaps check if the broadband noise generally decreases as you wear the headband for longer.

The absolute power bands and the FFT coefficients are logarithms of power. So if the systems you’ve used before provided linear-scale power, there will definitely be a noticeable difference.

Changes in alpha are most observable in the ear electrodes on Muse. Make sure that these are well connected and receiving good signal (the horseshoe indicator in the top right corner of MuseLab can help with this). Open up MuseLab and graph alpha_relative, closing your eyes for 10-20 seconds and opening them. You ought to be able to see some sort of change in the ear electrodes. There will still be short-term fluctuations given the short epoch we use. You can try smoothing to mitigate that.

You can actually smooth Muse data right in MuseLab using the following process (which is somewhat technical):

[B]Adding an averaging filter using MuseLab: [/B]

  1. Click on DSP Settings in the main dropdown menu.
  2. Click Create… beside DSP Settings
  3. Select the signals you want to filter (/muse/eeg 0-3) in the left list and move them to the right list. Click Finish.
  4. Beside Applied DSP Functions click Create…
  5. Click Filter and then Next.
  6. Type “0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1” into the input coefficients box and click Apply and then Finish. These coefficients give the average over ten samples. To change the size of the average to n samples, you need n coefficients of value 1/n. In this example you have 10 coefficients of value 1/10.
  7. Check the box beside the filter you’ve just created.
  8. Go back to the Visualizers tab using the dropdown menu. There will be new versions of the /muse/eeg signals with (dsp) beside them. Add them to the graph. They will be smoother than the originals.

Hope that helps!