Fourier transform for feature extraction of Electro Encephalo Graph (EEG) signals

Abstract

Electro Encephalo Graph (EEG) is a device that can capture electrical activity in the brain and inform the mind's condition such as emotional, fatigue, alertness, health and concentration level. Modelling the EEG signal before classification needs to be done, several studies have been carried out using Wavelet, Power Spectral, or Autoregressive transformations as feature extraction. This study explains the application of K-Nearest Neighbour as a classification and Fourier transform by taking the value of Power Spectral for feature extraction from the wave signal Electro Encephalo Graph (EEG). This study aims to identify EEG signals used for cursor movements. The data used are EEG data originating from the 2003 BCI competition (BCI 2003 Competition). The data contains data class 0 (for movement of the cursor up) and class 1 (for movement of the cursor down). Decision making is done in two stages. In the first stage, the Power Spectral value for each EEG signal is used to extract the feature. The feature is input to K-Nearest Neighbour. In the second stage of the identification process into two classes (class 0 and class 1) EEG signal data files, there are 250 EEG signal file training data and 25 from EEG signal file testing data, so that the total becomes 300 EEG signal data files. The results obtained for the classification of EEG signals are 84 % of the signal data tested.