Feature Extraction and Classification of Meditation EEG Signals

Abstract

The Meditation EEG signals are at the focus of scientific investigations due to the manifold benefits associated with it. The EEG signal analysis is implemented using wavelet decomposition and feature extraction. The Daubechies ‘db4’ wavelet is used for 6 level decomposition to obtain EEG sub bands. The statistical feature extraction based on wavelet coefficients is implemented. The combination of data set with inclusion of normalized coefficient band power values, as well, statistical features of wavelet coefficients such as maximum, minimum, average value, standard deviation, entropy, etc. are obtained. The ensemble subspace KNN classifier has a training accuracy 86.7 %. The testing accuracy has been 71.87 % to distinguish a EEG signal of a meditator before and after meditation.