ABSTRACT

Adaptive BCIs have shown promising results by reaching higher performances and robustness. However, existing methods lack guidelines in their manner of addressing the problem. Most of the current adaptive techniques simply handle challenges at hand by dynamically adjusting the machine to signal variability, often not entirely taking into account the human factors. To our knowledge, there has not been any work done in creating a taxonomy for adaptive BCIs, one that acknowledges and arranges known BCI components in a comprehensive and structured way. We propose a conceptual framework that encompasses adaptation approaches for both the user and the machine, that is, using instructional design observations as well as the usual machine learning techniques. This framework not only provides a coherent review of such extensive literature but also allows the readers to clearly visualize which component is being adapted and for what reason. Moreover, it enables the readers to perceive gaps in current BCIs 596and grasp the adaptive system in its entirety. Our proposal hopefully contributes as a guideline for a computational implementation of a fully adaptive BCI and an overall improvement.