mdh.sePublikationer
Ändra sökning
RefereraExporteraLänk till posten
Permanent länk

Direktlänk
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf
NSGA-II DESIGN FOR FEATURE SELECTION IN EEG CLASSIFICATION RELATED TO MOTOR IMAGERY
Mälardalens högskola, Akademin för innovation, design och teknik.
2020 (Engelska)Självständigt arbete på grundnivå (kandidatexamen), 10 poäng / 15 hpStudentuppsats (Examensarbete)
Abstract [en]

Feature selection is an important step regarding Electroencephalogram (EEG) classification, for a Brain-Computer Interface (BCI) systems, related to Motor Imagery (MI), due to large amount of features, and few samples. This makes the classification process computationally expensive, and limits the BCI systems real-time applicability. One solution to this problem, is to introduce a feature selection step, to reduce the number of features before classification. The problem that needs to be solved, is that by reducing the number of features, the classification accuracy suffers. Many studies propose Genetic Algorithms (GA), as solutions for feature selection problems, with Non-Dominated Sorting Genetic Algorithm II (NSGA-II) being one of the most widely used GAs in this regard. There are many different configurations applicable to GAs, specifically different combinations of individual representations, breeding operators, and objective functions. This study evaluates different combinations of representations, selection, and crossover operators, to see how different combinations perform regarding accuracy, and feature reduction, for EEG classification relating to MI. In total, 24 NSGA-II combinations were evaluated, combined with three different objective functions, on six subjects. Results shows that the breeding operators have little impact on both the average accuracy, and feature reduction. However, the individual representation, and objective function does, with a hierarchical, and an integer-based representation, achieved the most promising results regarding representations, while Pearson’s Correlation Feature Selection, combined with k-Nearest Neighbors, or Feature Reduction, obtained the most significant results regarding objective functions. These combinations were evaluated with five classifiers, where Linear Discriminant Analysis, Support Vector Machine (linear kernel), and Artificial Neural Network produced the highest, and most consistent accuracies. These results can help future studies develop their GAs, and selecting classifiers, regarding feature selection, in EEG-based MI classification, for BCI systems.

Ort, förlag, år, upplaga, sidor
2020. , s. 68
Nationell ämneskategori
Datorsystem
Identifikatorer
URN: urn:nbn:se:mdh:diva-47247OAI: oai:DiVA.org:mdh-47247DiVA, id: diva2:1413219
Ämne / kurs
Datavetenskap
Presentation
2020-01-31, 11:35 (Engelska)
Handledare
Examinatorer
Tillgänglig från: 2020-03-23 Skapad: 2020-03-09 Senast uppdaterad: 2020-03-23Bibliografiskt granskad

Open Access i DiVA

NSGA-II DESIGN FOR FEATURE SELECTION IN EEG CLASSIFICATION RELATED TO MOTOR IMAGERY(1715 kB)33 nedladdningar
Filinformation
Filnamn FULLTEXT01.pdfFilstorlek 1715 kBChecksumma SHA-512
159e40124cddb247359d9d08db0327f787b5048c6ab74c682186f2132c2e0b478c82bd4bab6555cfa277f84fd9d9906dc98774b7a4802616e25b5d42b736c661
Typ fulltextMimetyp application/pdf

Sök vidare i DiVA

Av författaren/redaktören
Johansson, Robin
Av organisationen
Akademin för innovation, design och teknik
Datorsystem

Sök vidare utanför DiVA

GoogleGoogle Scholar
Totalt: 33 nedladdningar
Antalet nedladdningar är summan av nedladdningar för alla fulltexter. Det kan inkludera t.ex tidigare versioner som nu inte längre är tillgängliga.

urn-nbn

Altmetricpoäng

urn-nbn
Totalt: 94 träffar
RefereraExporteraLänk till posten
Permanent länk

Direktlänk
Referera
Referensformat
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Annat format
Fler format
Språk
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Annat språk
Fler språk
Utmatningsformat
  • html
  • text
  • asciidoc
  • rtf