Blink, etc. are indistinguishable on the signal display.

[ilja.kuzovkin]

Hi!
Sorry for dumb question:

I have Emotiv EPOC.
When I visualize the signal from it with Aquisition server -> Signal Display, I get something like this (whatever I think or do) on every channel.

odd_signal.PNG (62.16 KiB) Viewed 20359 times

On the other hand if we look here http://common.ziffdavisinternet.com/uti ... 770,00.jpg we will see nice readings of some event.
They make much more sense than my output.

I tried to apply some filters as it is told in tutorials, but with no luck.
I'm new to all this, so may be I was doing something wrong.

So the question is - my output is OK and I just have to process it in the right way?
Or something is wrong with my device/emotiv driver/openvibe installation/brain ?

[yrenard]

Dear ilja kuzovkin

actually, the way the signal display works won't allow you to see eye blinks the same way as in the emotiv visualization widget. Indeed, the emotiv version allows signals to overlap each other. This behavior is not allowed by the current implementation of the signal display. So my feeling is that the portion of signal that you sent corresponds to the flat portions of the emotiv displays. When you blink, you should indeed see a peek, but this peek will cause the flat parts to be more flat.

I hop this was clear, tell me if it helps
Yann

[ilja.kuzovkin]

Yes, I understand what do You mean.
Like before the blink it will be

before.png (4.38 KiB) Viewed 20340 times

and after it will become something like

after.png (3.18 KiB) Viewed 20340 times

This is exactly the behavior I was expecting to see.

But here is a new sample, where I blink every 5th second (This is channel F8, other ones looks in the same way - no peeks)

blink5sec.PNG (106.54 KiB) Viewed 20340 times

Looks like it reads nothing at all, like then just lying on the table...

[yrenard]

Dear ilja.kuzovkin,

that actually looks like powerline interferences. Can you compute a spectral analysis on this signal and see if 50 Hz (or 60 Hz depending where you are based) is over represented ?

Hope this helps,
Yann

[ilja.kuzovkin]

Yes, it seems that 50 and 60 Hz are both presented

without-filter.PNG (97.78 KiB) Viewed 20333 times

I tried to apply Temporal filter like this

with-filter.PNG (139.25 KiB) Viewed 20333 times

But "blink" (or any other facial expression) still isn't visible...

( by the way - why despite Temporal filter is set to "3Hz to 24Hz" I still can see 25+ Hz on the spectrum display?)

[yrenard]

Dear ilja.kuzovkin,

50/60 Hz are not over represented. If over-represented, I expected a peak like the one you have at 0-1Hz on the first snapshot
Also, when you filter the signals using temporal filter, you can't remove the frequencies that are out of the specified range. This is a butterworth filter you are using so it basically tries to keep the signal within your range and attenuate what is out of the range. This is why you have this "gaussian" like distribution of frequencies. I suggest you read more about Butterworth filtering.

Now back on your initial concern not having the blinks in the signals ; is it possible that the channels are not correctly named for some reason ? Maybe you could show us the signals of all the channels when you blink ?

Yann

[ilja.kuzovkin]

50/60 Hz are not over represented. If over-represented, I expected a peak like the one you have at 0-1Hz on the first snapshot

Then may be this 0-1 Hz is responsible for my problem? Where it can come from?

Maybe you could show us the signals of all the channels when you blink?

Here are 2 new records. One is raw:
(Couldn't upload bigger images, channel names are AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, AF4 in same order)

5sec-blink-raw.jpg (212.82 KiB) Viewed 20310 times

and second is using Temporal filter (Butterworth, Band pass, 3 to 40 Hz):

5sec-blink-filtered.jpg (180.43 KiB) Viewed 20310 times

On both I blinked every 5th second.

[yrenard]

Dear ilja.kuzovkin,

the 0-1Hz peak comes from the fact that your signal is not centered on 0, so there is a constant component with a huge power.

The signals look like mainly noise. If you were blinking, you should definitely see them.

Can you give more details on you configuration ? (e.g. OS version, Emotiv SDK version etc...)

Yann

[ilja.kuzovkin]

Can you give more details on you configuration ? (e.g. OS version, Emotiv SDK version etc...)

I use Win7, edk.dll version 1.0.0.3, OpenViBE version 01-Oct-2010.

I also tried to use Emokit python library https://github.com/daeken/Emokit/blob/m ... ncement.md on linux.
It also shows no reaction on blinking...

Screenshot-pygame window.png (49.19 KiB) Viewed 20293 times

[lbonnet]

Dear ilja.kuzovkin,

I am using the same Emotiv with OpenViBE with no problem. But sometimes I do get the same noisy signal that you have here, but it is due to bad signal quality, especially on the 2 reference electrodes.

Do you check the signal quality with the Emotiv Control Panel before doing your measures ? (getting all dots green)
Does everything work on the Control Panel ?

PS: please don't use the emotiv control panel and openvibe acquisition at the same time !

[ilja.kuzovkin]

I am using the same Emotiv with OpenViBE with no problem.

This is gives me some hope...

I do it like so:
1. Put device on
2. Open Control Panel, make sure everything is green
3. Close Control Panel
4. Start acquisition

I tried to switch rubber ones with reference ones (as it is told to try in the EPOC manual: "swap the reference sensors with the rubber comfort pads").
But with no luck

[lbonnet]

I'm very confused...
Here is what I do in details :

1. Put device on
2. Open Control Panel, make sure everything is green
3. Close Control Panel

4. Start OpenViBE Acquisition Server
5. Set "Sample count per sent block" to 8 (the sampling frequency is 128Hz, so block of 8 samples ensure a smooth display - with 32 it would be clipped)
6. Connect
7. Play

8. Start OpenViBE Designer
9. Add a Acquisition client, hostname "localhost", port "1024"
10. Connect it to a Signal Display
11. Press play.

Then I see the EEG, and it responds well to blink (small spikes especially on AF electrodes) and jaw clenching (noise on all electrodes, especially Temporal ones).

If you don't have any problem on the Control Panel nor on openvibe consoles... I don't see where the problem is.

Did you tried the Control Panel functionalities ? The Cognitive suite and others, to see if they work as they should.

Did you contact Emotiv support to see if they can help ?

Laurent

[lbonnet]

Hey ilja,

Any update ? did you manage to get your EPOC working (either with the Emotiv Suite or OpenViBE) ?

Best Regards
Laurent

[ilja.kuzovkin]

Not yet
Yesterday tried on a almost bold person, same picture.
Going to speak with local EEG related person, maybe he will propose something.

[ddvlamin]

just a side note, but some say that the EEG of a bold person can be worse, because of a thicker skin. However, eye blinks should still be clearly visible.