BCILAB: a new tool for brain-computer interfacing

• ¹ University of California San Diego, United States
• ² University of California San Diego, CERCO, United States

BCILAB is a new open-source MATLAB-based toolbox and EEGLAB plugin aimed at helping advance the state of the art in designing and building Brain-Computer Interfaces (BCIs) and Cognitive Monitoring systems. These systems translate, in real time, brain and/or other biosignals into control outputs useful to computer applications, thereby offering communication and control pathways to their users that extend the brain's natural motor output channels. To address the challenges of accurately and robustly translating a person's brain signals into estimates of cognitive/affective state, and of solving the related inverse problems given the least amount of (relatively costly-to-acquire) data, BCILAB includes an extensible component framework for advanced signal processing, machine learning and related probabilistic inference, optimization and system evaluation. Among the provided methods are signal processing tools such as variants of Common Spatial Patterns, Independent Component Analysis, Time/Frequency Analysis, and machine learning tools such as Support Vector Machines, (sparse) logistic regression, variants of Gaussian mixture models and others, plus pre-defined standard BCI approaches constructed from them. The core of the toolbox is a framework that allows for the efficient GUI- or script-based prototyping, learning/calibration, offline testing, cross-validation, parameter search, visualization and real-time application of BCI designs, with support for cluster or local computation on Linux, Windows and Mac machines. In the near future, the toolbox will move further towards fully supporting mobile brain/body imaging (MoBI), which includes motion capture, eye tracking, and other measures, and towards more elaborate statistical data modeling, such as multi-task learning. Funding: Supported by The Swartz Foundation and US Army