Frequency analysis of signal acquired by SSVEP scenario

Working with OpenViBE signal processing scenarios and doing scenario/BCI design
Post Reply
lezan
Posts: 5
Joined: Sat Mar 23, 2013 10:21 pm

Frequency analysis of signal acquired by SSVEP scenario

Post by lezan »

Hi guys,
I need to check if I can found some SSVEP responses to a visual stimuli.
As a device to acquire the signal I used the MindSet, so I am aware of the fact that it is probably not possible to see nothing if the signal is acquired with this device, but I need to verify it. I have acquired the signal and I do not seem to see any kind of a response in the frequency domain analysis. The frequencies used are 12 and 15 Hz. Can you see if I'm right? Here you can obtain the signal acquired by me.
Moreover, I am a newbie signal analysis, but can you confirmed that it is useless a spatial filter CSP for a signal acquired by a device with only one electrode?
With a little editining to default SSVEP scenario (two target instead of three) I obtain this results

Code: Select all

[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 1 / 10 (performance : 48.1013%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 2 / 10 (performance : 44.9367%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 3 / 10 (performance : 51.5723%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 4 / 10 (performance : 36.0759%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 5 / 10 (performance : 42.7673%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 6 / 10 (performance : 50.6329%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 7 / 10 (performance : 47.4684%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 8 / 10 (performance : 47.1698%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 9 / 10 (performance : 49.3671%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Finished with partition 10 / 10 (performance : 49.6855%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.2> Classifier performance on whole set is 46.7777% (sigma = 4.35194%)

Code: Select all

[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> k-fold test could take quite a long time, be patient
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 1 / 10 (performance : 54.4304%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 2 / 10 (performance : 58.2278%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 3 / 10 (performance : 60.3774%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 4 / 10 (performance : 58.8608%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 5 / 10 (performance : 56.6038%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 6 / 10 (performance : 63.9241%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 7 / 10 (performance : 55.0633%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 8 / 10 (performance : 59.7484%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 9 / 10 (performance : 53.7975%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Finished with partition 10 / 10 (performance : 53.4591%)
[  INF  ] At time 298.117 sec <Box algorithm::Classifier for Frequency n.1> Classifier performance on whole set is 57.4492% (sigma = 3.21185%)
So, I realy have this accuracy?

Thanks for any answers.

jlegeny
Posts: 239
Joined: Tue Nov 02, 2010 8:51 am
Location: Mensia Technologies Paris FR
Contact:

Re: Frequency analysis of signal acquired by SSVEP scenario

Post by jlegeny »

Hi lezan,

sorry for a delayed reply. Indeed in your case there is no SSVEP response present in the signal. The accuracy of around 50% inside a binary classifier is consistent with chance level, i.e.: your classifier selects a class at random.

CSP itself is a spatial filter (linear combination of available channels) and thus needs at least two channels to work.

Regards
Jozef

Post Reply