Dispersion Entropy-based Fuzzy Classification for Motor Imagery Electroencephalogram


Balcı F.

International Conference on Artificial Intelligence and Applied Mathematics in Engineering (ICAIAME 2022), Baku, Azerbaijan, 20 - 22 May 2022, pp.1-3

  • Publication Type: Conference Paper / Summary Text
  • City: Baku
  • Country: Azerbaijan
  • Page Numbers: pp.1-3

Abstract

Fuzzy logic-based systems are a widely used method for both classification and control. Electroencephalography sign is frequently used in some studies such as Brain-Computer Interfaces. In this study, the popular and up-to-date Dispersion Entropy method is used to determine the values of fuzzy membership functions. The dynamic properties of time series are an important feature for classification. Some entropy methods (Sample Entropy, Permutation Entropy) are not successful in long time data. Therefore, noise bandwidth and simultaneous frequency-amplitude changes in the time series can be detected with the Dispersion Entropy method, which can be considered as new. The publicly available data set BCI:2000 was used to analyze the performance of the classification. In this data set, which was created with 14 basic tasks, 109 subjects have one or two minutes of EEG data. It is aimed to classify these tasks by focusing on 4 basic tasks. The usability of the algorithm designed with a high accuracy rate (92.3%) has been proven. In this way, it can be said that BCI systems can be used instead of systems with intensive processing costs such as deep learning.