Feature Extraction ElectroEncephaloGram (EEG) using wavelet transform for cursor movement

Abstract

This study aims to extract features related to human brain signals associated with ElectroEncephaloGraph (EEG) signal measurements and EEG signal classification extracted to relevant brain regions. EEG brain signals from 6 electrodes / channels placed on human scalp are recorded non-invasively using an EEG recorder with a 256 Hz sampling rate. The EEG data of the human brain function associated with the motor image task generated consists of two distinct activity classes, namely the Subject is asked to move the cursor up and down on the computer screen. The data used comes from the data set 1a BCI 2003 competition, consisting of two classes of class 0 as much as 135 experiments, 1st class of 133 experiments and trial data of 293 experiments. From the EEG signal data is processed using wavelet transform as feature extraction. Extract value of the characteristic value of the average, maximum, minimum and standard deviation. As for the classification using artificial neural network backpropagation. As a result, identification accuracy level using discrete Wavelet Transformation is 75%.