Binary classification of meditative state from the resting state using EEG

Objective classification of the meditative state of the brain from its resting state was attempted using electroencephalogram (EEG). The binary classification was performed both under intra-subject and inter-subject settings. The meditation practice used is Rajayoga meditation, which is probably the only meditation practice where the practitioners meditate with their eyes open. To our knowledge, this is the first attempt at such an EEG-based classification of altered conscious state during Rajayoga meditation. Baseline EEG was recorded both before and after the meditation session, separately with eyes closed and open. Common spatial pattern with Tikhonov regularization (TR-CSP) is used for feature extraction and linear discriminant analysis (LDA) is used as the classifier. We have achieved an accuracy of 97.9% for intra-subject classification and 74.0% for inter-subject classification. The classification accuracy improves with the increase in the number of filter pairs used; however, the improvement plateaus for all the subjects after about eight pairs of filters. Regularization by penalizing higher values of the elements in the filter vector deteriorates the accuracy and in most cases, the best classification occurs with no regularization.