Feature Selection in Electroencephalographic Signals Using a Multicriteria Decision Analysis Method

Abstract

In recent years, industry 4.0 has promoted the rise of new technologies and devices that generate and collect data both in industry and in everyday life. As a result, there is a new challenge of creating robust tools that facilitate the analysis of this information. Feature selection allows the efficient extraction of features that describe objects or phenomena, by removing distractors and redundant information. Thus, classification and decision-making models can be created with stable representations. In this chapter, a set of typical testors was extracted from a database with motor imagery EEG signals employing the testor theory, an approach to feature selection. A typical testor is the smallest possible combination of features that allow objects to be differentiated belonging to different classes. The purpose of this chapter is to select the typical testor with the best performance in an ANN classifier employing TOPSIS, a multicriteria decision analysis method.