AD-HOC GAUSSIAN DICTIONARIES FOR SPARSE REPRESENTATION OF EVOKED RELATED POTENTIALS

VICTORIA PETERSON; LEONARDO RUFINER; RUBÉN SPIES

Buenos Aires

Congreso; 1st Pan-American Congress on Computational Mechanics; 2015

CIMNE-MECON

A Brain-Computer Interface (BCI) is a system which provides direct communication between the mind of a person and the outside world by using only brain activity (EEG). A common EEG BCI paradigm is based on the so called Event-Related Potentials (ERP) which are responses of the brain to some external stimuli. For the present work at hand, the innermost part of a BCI is the pattern recognition stage whose aim is to detect the presence of ERPs with high accuracy. In recent years there has been a growing interest in the study of sparse representation of signals. Using a dictionary composed of prototype atoms, signals are written as linear combinations of just a few of those atoms. This sparse representation is found to be appropriate for posterior classification purposes. In this work we propose a sparse representation and posterior classification of ERPs signals by means of an ad-hoc spatio-temporal dictionary composed of bidimensional Gaussian atoms. Based on l1 -minimization we find the sparsest possible solution which allow us to design a robust classification based on nearest representation.