Statistical features extraction for multivariate pattern analysis in meditation EEG using PCA

This work was undertaken to study the specific statistical features of EEG data collected during meditation (Kriya Yoga) and normal conditions. The meditation practice changes the attentional allocation in the human brain to visualize this; statistical features are carefully calculated from different wavelet coefficients to categorize two diverse groups (i.e. Meditators and Non-Meditators). The entire time series of EEG data divided into overlapping segments, and statistical parameters calculated for each of these segments. Instead of using all the data points, we used only a few higher order statistical measures such as variance, kurtosis, relative band energy, Shannon entropy, and Renyi entropy obtained from the data segments. A standard clustering technique, i.e. Principal Component Analysis (PCA) used to get the distinct pattern from the statistical features in EEG. In this paper, we presented a clustering paradigm that used for the pattern analysis between meditators and non-meditators. We measured the EEG signal using 64 channels, with some peripheral physiological measures. 23 participants with varying experience in meditation practice and ten non-meditators (control group) are considered to visualize underlying clusters within the statistical features.