Online EEG Classification of Meditative States for Wearable Devices using Machine Learning

Wearable devices with EEG acquisition have been used to provide or assist meditative interventions. This paper presents a simulation scheme for wearable EEG devices for real time detection of mental states. Relative spectra and derived rhythm energies were used for classification features. Four machine learning algorithms, LDA, Bayes, SVM and KNN were applied to test 10 subjects who were instructed to perform 40 min focus-meditation. Offline training and testing were investigated to optimize the parameters of the classifiers, which yield an efficient computation method for the online model. The results demonstrate that the online detection pipeline is viable in real-time applications of wearable EEG devices.