A Bayesian Perspective on the Design of EEG-based Brain-Computer Interfaces for Augmentative and Alternative Communication and Control
Electroencephalography (EEG)-based brain-computer interfaces (BCIs) promise to provide a novel access channel for assistive technologies, including augmentative and alternative communication, and control of external devices for people with severe speech and physical impairments (SSPI). Research on the subject has been accelerating significantly in the last decade and the research community has taken great strides toward making EEG-based BCI a practical reality for individuals with SSPI. Nevertheless, there is much work to be done to produce viable systems that can be comfortably, conveniently, and reliably used by individuals with SSPI. This presentation will describe a Bayesian approach for the design of an EEG-based BCI. A dynamic graphical model will be developed to model the temporal dependencies in EEG, to design a decision fusion model for intent inference, and to design adaptive optimum stimuli subset selection schemes. RSVP KeyboardTM, a language model assisted EEG-based letter-by-letter typing system will be used as the testbed to demonstrate the design methodology.