Feature Subset Selection in Electroencephalographic Signals Using Typical Testors

Abstract

Motor imagery (MI) is a mental representation of movement without per-forming or tensing any muscles. MI requires a conscious activation of the same brain regions involved in actual movement. Brain signals have been explored for multiple applications in biomedical engineering, such as the development of brain-computer interfaces (BCI). BCI systems are designed to translate users' intentions into control signals, commands, or codes. Nevertheless, the major challenge in BCI system development is classifying MI signals recorded by an electroencephalogram (EEG). This paper focuses on applying the testor theory and the logical combinatorial pattern recognition approach for feature selection to reduce the feature representation space for classification tasks. The EMOTIV EPOC+ EEG device recorded the MI-EEG signals with 14 electrodes.