Revisão e comparação de classificadores em um experimento utilizando interface cérebro-computador

Abstract

A Brain-Computer Interface (BCI) can be used to control various electronics devices such as computers, smartphones and movements of a robot, for example. The construction of a BCI can be divided into five stages: signal acquisition, preprocessing, feature extraction, classification and application. For the classification stage classifiers are used to detect the user’s intentions that are performed in the desired application. However, there is a wide range of classifiers available, and not all classifiers can have a good performance for the desired application. In this context, the Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) classifiers were selected to be evaluated using an ICC application. The application used was P300 Speller, which is a speller implemented in OpenViBE software. Throughout the work, we describe the hardware for signal acquisition and also the scenarios implemented in OpenViBE software used to create the speller following the paradigm P300. Two experiments were performed denominated Case 1, wherein the LDA algorithm hit 7% of targets and Case 2, wherein the LDA and SVM algorithms hit 16% and 8%, respectively.