Can we called Muse "a real EEG"?


Can we called Muse “a real EEG”? In some Internet discussion I’ve read that devices like Mindwave from Neurosky in only a toy, which doesn’t read real brain waves but only galvanic skin responses (see: What do you think about this? Does the Muse read real brain waves?


So far I also suspect it.


Well … if galvanic skin responses EEG headbands can give a recording of two hours of sleeping with half an hour of a dreaming (a nightmare in fact, very scary dream!) showing very strange signals - starting with about 10 minutes of very high and irregular spikes only at the right forehead lobe (FP2), followed by another 15 minutes of the same kind of signals (that I was unable to get when awake) now totally synchronized at the four channels, ending only at FP2 again (and going back to normal - when I got awake) … then I think that Muse is a very good “galvanic” EEG headband :o

As a side note, when doing regular recordings with Muse when awake (both actively thinking and focused or when getting calm or trying to meditate) I allways get higher amplitude signals at the ears TP9 and TP10 compared to the sensors at the forehead FP1 and FP2. In that recording of a dreaming state all four channels went to much higher and all about the same amplitude - funny !

I believe that if galvanic skin potencials occurs, they should be considered as noise and I suppose that’s why the REF and DRL sensors are there, to supress them - it would be very nice if someone of the Muse Team could confirm or explain this point.

I will try to make a video of this EEG nightmare recording telling simultaneously what was happening (at least what I can remember) during the dream, very soon, and share if anyone interested.


BTW: before anyone ask me … yes, I enabled --dsp during recording, and checked when replaying all relevant indicators (is_good, quantization, horseshoe. all = 1 almost all the time, and mainly accelerometer showing no movement of the head - when it shows I disregard that part of the EEG.

As a toy to fool my nieces and nephews I developed a small windows program using only eye blinks and jaw clenches, asking them "… wich LED number do you want me to turn ON or OFF with my mind ? may be one at a time, or all ON or all OFF at once. " … they get crazy with the precise results :o


Muse is now the greatest of cheap eeg and probably could compete with some professional eeg. I’ve been using medical grade eeg for years now from different companies and muse sampling rate with its ease of use is marvelous. I’ve for first time seen clear signal up to 400-600Hz, it’s still full of artifacts but rest have good quality.


Muse is now the greatest of cheap eeg and probably could compete with some professional eeg. I’ve been using medical grade eeg for years now from different companies and muse sampling rate with its ease of use is marvelous. I’ve for first time seen clear signal up to 400-600Hz, it’s still full of artifacts but rest have good quality.
Y for Hz and X for time…seFFT.png?dl=0


[B]Eduardo, could you share code of your app?[/B]


Hi there,

Thanks for your interest and curiosity!

I can assure you that Muse does in fact measure actual EEG signals. It does this by measuring changes in the electric field near the head as they manifest as voltages on the surface of the scalp. Each channel uses a very sensitive voltage amplifier. Of course what this means is that anything that affects the voltage at a certain point on the scalp affects the signal that Muse picks up. So in addition to EEG (brainwaves), Muse can sense EMG (muscle activity, e.g. clenching your jaw or furrowing your brow), EOG (changes in the strength and direction of the corneoretinal potential of the eye, e.g. eye movement, blinks), ECG (your pulse - although you would have to position the electrodes on your chest and arms or place one of your fingers on an electrode and the fingers your other hand on DRL and REF to see it clearly), and yes, the effects of GSR (after all, changes in skin conductance will certainly affect the skin voltage).

However, GSR is a fairly slow-changing signal. It does not vary as rapidly as EEG signals, so it’s in a different part of the frequency spectrum than they are, and so the two very rarely interfere. It’s generally quite easy for us to tell the difference between EEG and signals that might be due to GSR.

Furthermore, we’ve actually performed tests comparing Muse’s signal quality and performance to that of a clinical-grade EEG system (the BrainVision actiChamp/actiCAP system) which showed that they were quite similar. See the attached images for more info. One is a spectral comparison, and one is a time-domain comparison, both with electrodes from Muse and the actiChamp recording simultaneously from the same location. Note how well they compare.

Actually, in that thread that you referred to, @nekrodezynfekator, the general consensus seems to be that the toy in question almost certainly does measure EEG, not GSR. The posters there make several good points to that effect, and even reference a study that compared that technology to a clinical-grade system.


Eduardo, how did you record/measure your dream experience to get this kind of data? My friend and I are interested in exploring the idea of inter and intra whole brain coherance. It’s one thing to relax by entering an alpha state, it’s another to develop coherance IN that alpha state of Beta, Theta and even Delta frequency mental activity–whole brain coherance. If you’re familiar with Compressed Spectral Array data (COSPARs) that characterize different states of consciousness, it seems like Muse should be able to do the same thing IF each sensor can be recorded and displayed separately portraying levels of frequency.
Does your software do this? Is it WIN or iOS? Is there anything like this out there?


Hi DaleTreece,

To answer your question about getting frequency information from each channel individually, that’s definitely possible. MuseIO outputs raw FFT coefficients for each channel as well as Alpha, Beta, Delta, Theta, and Gamma measures. More info here:

Looking forward to hearing more about your and Eduardo’s projects, too!


rudzinskimaciej, can you tell me what software and FFT/other parameters you are using to generate the spectrogram in your post? I’m looking to understand best practices on this (see this post for more info:…hon-matplotlib )

[Edit: I’m now able to create spectrograms that are quieter at lower frequencies, by subtracting the mean of the EEG data prior to computing the FFT. Thanks to Tom for pointing this out].


Eduardo, please could you share code of your toy app Muse Home Automation Control ?


I made my version (Muse Home Automation Control)


Please post images comparing the latest Muse EEG with a reference EEG headset. This is an old post and the comparison pictures are missing. I’d like to see the comparison and especially with the latest Muse model.


Here you go:

You’ll note that this is an open source project, with all the relevant info here:

You can also try Alexandre Barachant’s MuseLSL work:


do you have articles about the comparison between muse and eeg clinical grade system?


Yes, that article above is a comparison between Muse and a research grade EEG. Research grade EEG is higher than clinical grade in terms of signal quality.