Genetic algorithm and forward method for feature selection in EEG feature space

摘要: There are a lot of problems that arise in the process building brain-computer interface based on electroencephalographic signals (EEG). A huge imbalance between number experiments possible to conduct and size feature space, containing features extracted from recorded signals, is one them. To reduce this imbalance, it necessary apply methods for selection. One approaches selection, often taken researches, classic genetic algorithm codes all within each individual. In study, there will be shown, although approach al- lows obtaining set high classification precision, also leads highly redundant comparing selected using forward selection method or equipped with individuals given (very small) genes.