Abstract:
Two artificial neural network systems were designed by using wavelet based
features for the classification of normal and abnormal EEG signals were
decomposed to 4 levels using Daubechies wavelet of order 2.These EEG signals
were decomposed to four statistical features: minimum, maximum, mean and
standard deviation to depict their distribution. These features computed over the
wavelet coefficients for each level, and used as input to the artificial neural
network systems. After training and testing the systems results were obtained for
classification of signals. The Two type of neural networks(Feed Forward Back
propagation and Cascade Forward Back propagation) were tested for sensitivity,
specificity and accuracy it was found that the Cascade Forward Back propagation
(CFBP) give more accurate results with an accuracy of 96.76%.
