M. Nuh, A. Jazidie, M.A. Muslim
Detecting and classifying sharp transients in EEG (Electroencephalograph) recording by visual screening is a laborious and time-consuming task. That is why, there is an urgent need to construct a computer algorithm to detect automatically that type of EEG transient phenomena. The use of an artificial neural network as a classifier and wavelet analysis as pre-processing give promising results to answer that need. This paper proposes to develop a new method for the automatic detection of epileptic spikes based on Wavelet Neural Networks (WNN). A proper selection of scaling in WNN is introduced to overcome the problem of very long time duration during training. The result shows that proper selection of wavelet scaling can decrease training duration without decreasing WNN performance. © 2002 IEEE.
Electronic Engineering Polytechnic Institute of Surabaya (EEPIS), Kampus ITS, Sukolilo, Surabaya, Indonesia; Electrical Engineering Department, Sepuluh Nopember Institute of Technology, Kampus ITS, Sukolilo, Surabaya, Indonesia; Electrical Engineering Dept., Brawijaya University, Jl. Mayjen Haryono 167, Malang, Indonesia
