Selección de parámetros en el enfoque de bolsa de características para clasificación de habla imaginada en electroencefalogramas

The use of brain-computer interfaces as a new communication channel has become an interest topic, nevertheless, there are many challenges to achieve a natural communication. In the particular case of imagined speech BCIs, there are difficulties to analyze the brain signals. In this work, a bag of features representation to identify patterns in electroencephalograms is explored. This method has shown to be useful in similar tasks. Nevertheless, to determine the most adequate bag of features to the problem depends on many parameters. The present work describes the use of genetic algorithms to find an adequate configuration to imagined speech classification. The results, obtained from an imagined speech database of 27 subjects, show the relevance of the method achieving results which are comparable with related work.