Matching pursuit based indices for examining physiological differences of meditators and non-meditators: An HRV study

Abstract

Since meditation has recently been used as a complementary therapy, the study of physiological effects of different kinds of meditation has been augmented over the last decade. In this study, a robust method based on the matching pursuit (MP) algorithm was proposed to examine the heart rate variability (HRV) differences of meditators and non-meditators. The non-meditators were comprised of metronomic breathing (MB) and spontaneous nocturnal breathing (SNB) groups. The meditators were comprised of pre Chinese Chi meditation (pCCM), during CCM (dCCM), pre Kundalini yoga meditation (pKYM), and during KYM (dKYM). Following MP based decomposition of HRV into its sparse parts, eleven indices were extracted for the MP coefficients. The efficiency of the indices was also examined through statistical significance between the groups. It was shown that for almost all indices, significant differences between the classes were observed. Furthermore, a probabilistic neural network (PNN) with a variable sigma (σ) value, was trained to classify the physiological responses of the groups. A maximum accuracy of 99.61% was detected for dKYM using 5-fold cross validation scheme. It was also shown that lower σ values give higher classification rates and upper σ values provide more robust accuracies. In conclusion, the proposed method is a promising technique for showing significant differences between HRV responses of different non-meditator and meditator groups.