Classification of Electroencephalographic Changes in Meditation and Rest: using Correlation Dimension and Wavelet Coefficients
International Journal of Information Technology and Computer Science March 27, 2012 Atefeh Goshvarpour, Ateke Goshvarpour 10 citations
An algorithm that distinguishes between resting and meditative states using electroencephalogram (EEG) signals was tested. EEG data from 25 healthy women were collected before and during meditation. Wavelet coefficients and correlation dimensions from electrodes Fz, Cz, and Pz were used as features for several classifiers. The Fisher discriminant and Parzen classifier achieved the highest accuracies: 85.02% and 84.75% with Wavelet coefficients, and 92.37% for both when using correlation dimensions. The findings suggest that nonlinear measures like correlation dimension are more effective than wavelet features for classifying meditation versus rest from EEG signals.