Spike and epileptic seizure detection using wavelet packet transform based on approximate entropy and energy with artificial neural network

This paper proposes the method that can detect both spikes and epileptic seizure at the same time based on wavelet packet transform, approximate entropy and energy, and artificial neural network. First, the EEG signals are decomposed into 4 levels, 16 frequency sub-bands, using Daubechies for mother wavelet to distinguish the usable signal. Then the approximate entropy and energy features are extracted for each sub-band to form the feature vector. Finally, the constructed feature vector is used as an input to the artificial neural network to classify the EEG signals into 6 types of spike, epileptic seizure, eye closed, eye opened, body movement, and normal signal. The experimental results show that the proposed method identified and classified the EEG signal with average sensitivity of 76.55%, specificity of 81.3%, and accuracy of 89.47%.