Abstract: Epilepsy is the most widely recognized chronic neurological diseases and the most well-known neurological chronic disease of childhood. The Electroencephalogram (EEG) signal furnishes significant information neurologists contemplate in the investigation and analysis of epileptic seizures. EEG being the non-stationary signals, legitimate analysis is greatly needed to classify the normal, inter-ictal and ictal EEGs. This srudy presents a propose method is to classify the EEG signals regarding the existence or absence of seizure. Modified EELM is more accurate for automatic epileptic seizure detection. Multi wavelet transform contains both scaling and wavelet functions, simultaneously offers orthogonality and symmetry which is not possible for scalar wavelet transform. With these major properties, the multi wavelet transform is promising in signal processing applications. Multi wavelet based characteristic is utilized to differentiate the normal EEGs of healthy subjects, inter-ictal and ictal EEGs of epileptic patients. To acquire the detail and approximation wavelet coefficients, the EEG signals are decomposed into sub-groups. The proposed strategy decomposes with wavelet transform, reduce dimensionality by a set of features and have a classification with Modified Effective Extreme Learning Machine (Modified-EELM). The modified-EELM has a better accuracy than ELM. The Modified-EELM Classification algorithm with the sigmoid function has 98.01% testing accuracy, good performance, easy to implementing and consumes only 0.0008 sec of time. This is very less amount of time compared with other learning machines.
A.S. Muthanantha Murugavel and S. Ramakrishnan, 2014. An Improved Detection of Epilepsy EEG Using Multi Wavelet Transform and Modified Effective Extreme Learning Machine. Asian Journal of Information Technology, 13: 639-645.