Epileptic Seizure Detection using Multicolumn Convolutional Neural Network

This paper attempts to utilize the Convolutional Neural Network (CNN) classifier by applying the preprocessing steps on the dataset of the EEG signals. Dividing the samples into separate frequency bands, the power spectrum for each frequency band is calculated. All the parameters are taken for calculating each discrete wavelet transform. By amalgamation of the features like wavelet statistics, the frequency band power and total power, a complete feature vector of a single set is derived. The average of the CNN columns is then concatenated with the 4 level Discrete Wavelet Transform operated on the Pre-Processed EEG Signal with Power spectrum calculation value for each frequency band by using machine learning techniques.This creates the Multi-Column Convolutional Neural Network which provides a 99.99% recognition rate.