A General Architecture for a Co-Learning of Brain Computer Interfaces

Traditionally, BCI research has been focussed on the signal processing and medical aspects of BCIs, while the aspects pertaining to interaction, usability and convenience, have been studied more scarcely. Commonly, training sessions are slow and tireing.

In the context of the CA-ICO (Co-apprentissage/Interfaces Cerveau Ordinateur, or in English Co-learning/Brain Computer Interfaces) project of LIG-IIHM in collaboration with GIPSA-lab and funded by the Grenoble INP Univeristy and of my thesis research, I am working towards putting co-learning between the system and the user at the center of BCI system design. The aim is to minimize offline training phases and maximize the user experience of BCIs. The ultimate goal is to bring BCI systems outside of the lab with a performance level comparable to more traditionnal and robust interaction modalities.

In recent work for the EMBS NER 2013 conference, I have proposed (jointly with Franck Tarpin-Bernard) the idea of a software architecture for asynchronous BCIs based on co-learning, where the system and the user jointly learn by providing feedback to one another.
The architecture relies on the use of recent filtering techniques such as Riemann Geometry and ICA, followed by multiple classifications with incremental classifiers.

Furthermore, I have now started implementing the first components of the architecture and performing preliminary experiments to first validate the general idea both quantitatively and qualitatively (user experience). The results indicate a good potential and are currently being prepared for publication in the near future.

For more information or any question or querries you might have, do not hesitate to contact me.

© Copyright 2015 Nataliya Kosmyna. All Rights Reserved.