UC San Diego

EEG-based brain computer interfaces (BCI) allow users to communicate with the outside world directly through signals read from the cortex. These systems typically require user training to generate signals appropriate for use. However, user ability to control such a system is variable and the variables involved are not fully understood. In this work, we present the foundation for the identification and use of feedback-related EEG signals to both augment existing motor imagery BCIs as well as to provide a novel mechanism for control. We conducted a simulated online experiment where users were instructed to move a cursor in one dimension from the center of the screen towards a target located at the left or right extremes. In Chapter 2 we demonstrate how local aspects of visual feedback (on a single cursor movement scale) can augment existing systems or provide control signals for novel ones. In Chapter 3, we demonstrate how global aspects of visual feedback can affect motor imagery performance. Rather than more positive feedback valence providing a more stable signal for motor imagery classification (as originally thought), feedback valence instead appears to rotate the motor imagery feature space. In Chapter 4 we validate our findings from Chapter 2 in an online, real-time setting. In addition, we demonstrate the utility of a BCI system that utilizes local aspects of user satisfaction and dissatisfaction as a principal control signal. This proves to be a useful, stable system that provides improvement over a traditional motor imagery paradigm