Classification of Epileptic EEG Signals using Time-Delay Neural Networks and Probabilistic Neural Networks

• DOI: 10.5815/ijieeb.2013.01.07
• Publication Date: 2013-05-28T05:13:26Z
• Authors: Ateke Goshvarpour; Hossein Ebrahimne...; Atefeh Goshvarpour
• Keywords: Approximate entropy

The aim of this paper is to investigate the performance time delay neural networks (TDNNs) and probabilistic (PNNs) trained with nonlinear features (Lyapunov exponents Entropy) on electroencephalogram signals (EEG) in a specific pathological state.For purpose, two types EEG (normal partial epilepsy) are analyzed.To evaluate classifiers, mean square error (MSE) elapsed each classifier examined.The results show that TDNN 12 neurons hidden layer result lower MSE training about 19.69 second.According results, when sigma values than 0.56, best proposed network structure achieved.The present study applying train these can serve as useful tool classifying signals.