New classification algorithm in OpenVibe Gipsa Extensions

[toncho11]

Hello,

We have added a new classification algorithm called MDM based on Riemannian Geometry. It supports multi-class naturally (compared to current implementations in OpenVibe). Implementation is in C++/Python in the form of OpenVibe boxes. Also we added scenarios that use the MDM for P300 and motor imagery. The adaptive version of P300 with MDM for Brain Invaders works without training. The algorithm can be used for SSVP also. More on MDM: A. Barachant, S. Bonnet, M. Congedo and C. Jutten, 'Multiclass Brain-Computer Interface Classification by Riemannian Geometry,'.

There are also two new boxes for resampling and merging of data streams in OpenVibe.

Our last version is 1.9:
http://code.google.com/p/openvibe-gipsa ... /Downloads
Release 1.9 is compatible only with OpenVibe 0.16.x.

1.9 has also the latest version of the P300 game Brain Invaders.

Happy summer hacking,
Anton

p.s. currently it is not possible to play BI with software tagging without recompilation and changing OpenVibe 16.1

[toncho11]

Note that the Adaptive MDM (no training) requires a certain electrode montage: http://openvibe-gipsa-extensions.google ... ontage.png
We are also testing the OpenVibe SSVP scenario with our MDM boxes, because in this scenario there are 4 classes.

[toncho11]

Version 1.9.6 has been released.

Emotiv scenarios has been added and BI can be used with software tagging (that is required for Emotiv).

The Brain Invaders has an improved launcher with an extra tab "Advanced" where you can just select Emotiv from the dropdown.