A New Classification Method of Epileptic Eeg Signals Using Differential Evolution Optimally Pruned Extreme Learning Machine

An epileptic seizure is a transient event of symptoms due to abnormal neuronal action in the brain. Electroencephalography (EEG) is the neuro physiological measurement of electrical activity in the brain as recorded by electrodes placed in the cerebral cortex. An epilepsy EEG is based on three approaches. First, a scaling and wavelet function of the Multi Wavelet Transform (MWT) offers orthogonality and symmetry. Second, Feature Extraction reduces the dimensionality. Finally, Extreme Learning Machine (ELM) is utilized to train a single hidden layer of feed forward neural network (SLFN) features for classification of EEG signal. In this paper, a proposed method introduces a machine learning algorithm referred to as Differential Evolution Optimally Pruned Extreme Learning Machine (DE-OPELM). The DE-OPELM is utilized to detect and classify the epileptic EEG in the diagnosis of patients with epilepsy. The DE-OPELM is applicable to high-dimensional complex optimization problems. Hence, the DE-OPELM is accurate than ELM for automatic epileptic seizure detection. For inter-ictal EEG and ictal EEGs, the DE-OPELM produces 98.33% testing accuracy. As compared with other learning machines, the DE-OPELM detects only 0.0007sec of time.