Abstract
The need and importance of automatic disease recognition such as electroencephalogram has grown over time. Driven by this motivation, this research demonstrates the electroencephalogram (EEG) signals reconstruction process using the finite impulse response, principle component analysis, feature extraction, and support vector machine (SVM). After mentioning related literature, the EEG signals are taken from biomedical database such as Temple University Hospital, Australia. Applying finite impulse filter to the noisy EEG signals, the motion artifacts have been effectively removed. Generally, EEG signal is a multidimensional so it is quite difficult to find out effective channel for different diseases. Applying principle component analysis over filtered EEG signals, dimensional reduced EEG signals are obtained. For classifying EEG signals, different statistical measured such as standard deviation and mean absolute deviation are applied. Moreover, the SVM is used to classify the EEG signal from the selected features. Finally, the system performance is evaluated by 27 patients EEG database. For each disease, it has taken 9 signals. For different signals, the SVM are trained and evaluate the performance. Simulation results show that the SVM provides better performance for higher number of signals.