Motor Imagery with CSP - Changing the Output

[peterold]

Hi,
I'm new to openvibe, and I very recently got everything up and running. I'm using the Emotiv headset with the Motor Imagery CSP, and so far I've been reaching over 80% accuracy. What I'd like to do now is to put something in that measures the accuracy, compares it to a threshold (say 80%), and if it's over 80% output = yes if not output = no. Is there something already available that can do that?

To expand upon the big picture: I'm doing a student design project to create an EEG password system. I'm looking to revamp the motor imagery CSP so that once you train the classifier you can run different inputs through it. If the new input matches the one that's been trained into the classifier then it lets you through. If not then you're denied.

All help, pointers, and tips are appreciated!
Thank you.

[jtlindgren]

Hi Peterold,

I'm sure there are many ways to achieve this. One of the simplest ones would be to write the classifier outputs to a disk (there are boxes that should be able to do this) and then launch your own application/script to compare the results against a stored pattern (there's a box for launching as well) and let your own app decide if the match is close enough.


Happy hacking,
Jussi

[peterold]

Hi Jussi,

I tried outputting it directly to a CSV file, but it just gave me three lines of useless information (time, identifier, and duration). Neither identifier or duration changed down the column. What I'd really like to do is to output the percent match shown by the classifier accuracy measure box but that doesn't have an output. I was looking into the voting classifier, but I don't really understand how it works even after I read up on it. I'm not sure if it would even be useful here.

Any hints?

Thank you,
-Peter

[jtlindgren]

If you connect a keyboard stimulator to CSV file writer (after changing its input to 'stimulations'), you should see that the keys you press get tagged in the CSV file with the corresponding, different identifiers. You should be able to get the same result by directing the stimulation output from the classifier to the CSV writer. In OpenViBE, the classifiers are typically two class, they output from among just a few configurable labels. If you wish to tell several classifiers apart, you need to make them output different labels. Then, you can use stimulation multiplexer to combine the various stimulus streams with different labels into a single stream and feed that to the CSV writer.


Cheers,
Jussi