I've recently got a Neurosky Mindwave Mobile and started trying to use it with OpenViBE. In particular I'm trying to use the SSVEP Mind Shooter scenarios.
After a bit of tinkering I've got the headset talking to OpenViBE and have been able to run through the Mind Shooter recording and training scenarios. When I run scenario 5 to play the game though, the performance is disappointing. In trying to figure out why, I've noted that the scenario says "Note: The results of the classifier are only used if the SSVEP_AdvancedControl is set to false".
I've hunted high and low through the box layouts, scenario documentation, config files and forums and can't find anything labelled "SSVEP_AdvancedControl". Has the control been renamed, changed or removed, or am I missing something?
Mind Shooter - How to use with trained classifiers?
-
jtlindgren
- Posts: 775
- Joined: Tue Dec 04, 2012 3:53 pm
- Location: INRIA Rennes, FRANCE
Re: Mind Shooter - How to use with trained classifiers?
Hi,
SSVEP_AdvancedControl is a boolean token you can put to openvibe.conf for example, the C++ code of mind shooter refers to it.
As far as I understand, the difference between the two controls is that the AdvancedControl bases the ship movements on the continuous-valued predictions of the classifier, for example class probabilities or distance from the hyperplane (either of the two green output sockets of the classifier processor box), whereas the other control type uses the discrete class label predictions from the classifier directly (the purple socket), but not their probabilities.
It may be challenging to get SSVEP to work with only a single electrode of the neurosky (and this electrode should be placed at the back of the head! not in the front), but if you manage to do it, we'd be delighted to know.
Cheers,
Jussi
SSVEP_AdvancedControl is a boolean token you can put to openvibe.conf for example, the C++ code of mind shooter refers to it.
As far as I understand, the difference between the two controls is that the AdvancedControl bases the ship movements on the continuous-valued predictions of the classifier, for example class probabilities or distance from the hyperplane (either of the two green output sockets of the classifier processor box), whereas the other control type uses the discrete class label predictions from the classifier directly (the purple socket), but not their probabilities.
It may be challenging to get SSVEP to work with only a single electrode of the neurosky (and this electrode should be placed at the back of the head! not in the front), but if you manage to do it, we'd be delighted to know.
Cheers,
Jussi
Re: Mind Shooter - How to use with trained classifiers?
Thanks.
I've added the SSVEP_AdvancedControl = false line to the application-configuration.conf file in the example folder. It does seem to have changed the behaviour. It's still not working well, but as you suggest that's likely because the Mindwave isn't well suited (and I'm still trying to use the electrode front-mounted - I don't think it'll work through the hair on the back of my head). I might try the reverse mount next time I get a haircut!
I've added the SSVEP_AdvancedControl = false line to the application-configuration.conf file in the example folder. It does seem to have changed the behaviour. It's still not working well, but as you suggest that's likely because the Mindwave isn't well suited (and I'm still trying to use the electrode front-mounted - I don't think it'll work through the hair on the back of my head). I might try the reverse mount next time I get a haircut!