Mental Tasks for Controlling Cursor Movements in a Brain-Computer Interface

Traumatic brain injury and neuro-degenerative diseases leave people dependent and with a low quality of life. Several technologies have been proposed to connect the central nervous system to silicon circuitry and thereby circumvent the damaged nervous tissue that is impairing normal function. These technologies rely upon a language of subtle movements within a patients capability to direct computer operations, however, selecting the proper abstraction for even a task as simple as moving a computer cursor can be challenging. Involuntary movements can create noise and false positives for the brain-computer-interface (BCI) receptors, and non-intuitive abstractions can be a barrier for adoption by neurologically damaged patients. We therefore introduce Visualization of Arrow Movements (VAM) as a set of mental tasks for controlling cursor movements in a BCI system. The performance of VAM was evaluated by six untrained subjects via 10-fold cross validation using band power and k-Nearest Neighbor classification methods as well as linear discriminant analysis (LDA) after spatial filtering. The binary classification accuracy in recognizing VAM tasks from each other was between 92% and 100% for four subjects and between 66% and 72% for the other two participants, which suggests that the tasks are most intuitive for even untrained persons. Non-EEG analysis revealed that this performance does not originate from ocular or other facial movements, but from cerebral electrical activity. The high classification accuracy and intuitive abstraction suggest that VAM is a promising abstraction for BCI systems.