THINK: Toward Practical General-Purpose Brain-Computer Communication

Abstract

In this work, we present THINK, a practical general-purpose brain-computer communication platform that relies on the OpenBCI and OpenViBE hardware and software platforms, and allows for a simple three-alphabet vocabulary. Specifically, we consider the scenario where a subject is wearing a sensor array (an electrode cap), and consciously manipulating her thoughts to communicate wirelessly with an external computing entity (a smartphone) without the aid of any external stimuli. Using THINK, we explore general aspects of brain computer communication that are application agnostic. In particular, we study the system accuracy and usability with real user experiments. The system accuracy was found to be highly variable across subjects and trials. We achieved a maximum accuracy of 83.4% and average accuracy of 53.4%. Even with low accuracy, we demonstrate that how is it possible to construct a successful BCC system. Further, in usability, we explore (i) how fast can the subject switch thoughts corresponding to symbols; (ii) is there an impact on accuracy with learning time; and (iv) how does accuracy drop with decreasing number of sensors (electrodes)? Using purely experimental analysis, we present some results that provide preliminary answers for these questions. We also provide motivation results for the future work in the context of (i) alphabet design as per user preference, and (ii) importance of pre-processing and requirement of better algorithms.