Eventos Anais de eventos
COBEM 2017
24th ABCM International Congress of Mechanical Engineering
Classification of EEG Signals using Genetic Programming
Submission Author:
Felipe Rebelo Lopes , RJ
Co-Authors:
Marco Antonio Meggiolaro, Felipe Rebelo Lopes
Presenter: Felipe Rebelo Lopes
doi://10.26678/ABCM.COBEM2017.COB17-1839
Abstract
Brain Computer Interface systems are based on an analysis of electroencephalogram signals, associated with an intention of a human being only based on thought. Feature extraction and classification are still a challenge. This paper presents the genetic programming technique as an alternative to pattern classifiers. Publicly available BCI competition IV dataset I, a multichannel 2-class motor-imagery dataset, is used for this purpose. The Wavelets Transform method is applied to decompose the signal in frequency sub-bands. In addition, different features are calculated. Measurement of time, statistics, and information theory are the most important features of this type of signal. Moreover, to classify features into two classes (imaginary movement of the right hand or feet), a multi-gene genetic programming is used to remove noise. It is shown that the training performance reaches a minimal error in 20 generations. The model is validated from its high rate of successfully classified commands.
Keywords
BCI Systems, Wavelet Transform, EEG Signals, Genetic Programming
