The value of the standard deviation of wavelet subband coefficients as feature extraction for electro encephalo graph (EEG) signal

Abstract

In principle, this study describes the classification of EEG signals using Backpropagation Neural Network as a classification and Discrete Wavelet Transformation for feature extraction by taking the standard deviation value on each Wavelet subband. The purpose of this research was to identify the EEG signals used in the movement of the cursor. BCI competition 2003 data sets Ia as the data for this research. This data contains the data classes 0 (for the movement of the cursor to the top) and class 1 (for the movement of the cursor to the bottom). EEG signals are classified in two stages. In the first stage, the value of the standard deviation on each discrete wavelet subband used to extract the features of EEG signal data. This feature as inputs on Back propagation Neural Network. On the second stage of the process into two classes (class 0 and class 1) EEG signal data file, there are 260 training data file 293 of EEG and signal data file signals EEG testing, so that the whole be 553 file data signals EEG. The results obtained for the EEG signal classification was 78.7% of the data signal is tested.