OPERATING VIRTUAL KEYBOARD WITH EEG SIGNALS

BCIs provide a means of communication for users with severe motor impairments, such as due to ALS or locked-in syndrome, through the decoding of their neural signals. Electroencephalography, or EEG, is considered a non invasive and broadly used BCI modality in which the power in certain frequency bands represents different cognitive or motor states. We have developed in this thesis an EEG-based virtual keyboard that accepts three intentional brain commands, namely, leftward, rightward, and eye-blink, to navigate a 5x6 on screen key matrix. Features of EEG bandpower in the alpha, beta, and theta bands extracted from scalp recordings served as input to machine learning classifiers. For example, prior work has shown that alpha/beta power can be highly predictive in EEG classification tasks and that eye blinks can be harnessed as a control signal to improve communication. In summary, the project illustrates that spectral EEG features along with ML classifiers can implement a brain-controlled virtual keyboard, pointing toward future assistive applications in augmentative communication and device control.