The best artificial neural network parameters for electroencephalogram classification based on discrete wavelettransform

Abstract

This paper presents the classification of electroencephalogram (EEG) signals using artificial neural network techniques. The signal processing of EEG signal could provide several areas for research in biomedical field. Numerous techniques can be applied to extract out the EEG characteristics in order to study and investigates the problems in the pattern recognition by its features extracted. The interesting site of signal measurement is the temporal lobe which is responsible of T3 and T4 in human electrode placement scalp. In this paper, many subjects were used to test the performance of non-neurophysiologic signals in order to investigate the electrical waves in human brain via the production of numerous EEG signals. A linear method of discrete wavelet transform (DWT) was used to gain classification with accuracy of 94.93% for testing EEG of different samples of music such as rock, jazz, classical and heavy metal using artificial neural network (ANN) with 2000 epoch, 25 nodes, 2 hidden layers. The results showed promisingly valuable EEG signal characteristics which could support the hospital staff to take care of and treat patients in the correct direction.