Extraction and recognition of electroencephalogram dynamic patterns for brain-computer interfaces

Abstract

We discuss an original approach to multidimensional non-stationary time series classification based on dynamic patterns analysis. The main problem in time series classification is construction of appropriate feature space. The success of classification dramatically depends on the quality of the feature space chosen. To construct this space we develop the method for extraction of dynamic patterns that are the most specific for the time series of each class. This problem is formulated as an optimization problem and the genetic algorithms are used to resolve it. The simulation results are given for the real electroencephalogram signals that are used in the brain-computer interfaces.