Title: "Signal Processing and Machine-learning Algorithms for Brain-Computer Interfaces."

Leonard J. Trejo

Roman Rosipal

NASA Ames Research Center

Abstract:

We are developing methods that draw control signals for human-computer interfaces from the human nervous system. We have made progress in three areas: a) experimental designs for EEG-based brain computer interface (BCI) development, b) identification of EEG sources using spatial filters and adaptive time-series filters, c) dimension-reduction of complex spatio-temporal EEG patterns using linear partial least squares (PLS) decompositions, d)classification of mental states from EEG signals using support vector machines (SVM)and a combined PLS-SVC method. We will give an overview of our research and discuss the remaining challenges in BCI development.