Kernel-based SVM (k-SVM) surpassed conventional SVM in classifying meditation and non-meditation allied EEG, achieving an average classification accuracy of 90.8% compared to 85.5%.
Does k-SVM improve classification accuracy of meditation state-allied EEG compared to conventional SVM?
Kernel-based SVM outperforms conventional SVM in classifying EEG signals associated with Kriya Yoga meditation.
Absolute Event Rate: 90.8259% vs 85.543%
Support vector machines (SVM) have become a gold standard method for the classification of brain signals. However, for highly nonlinear and non-stationary signals like Electroencephalography (EEG), conventional SVM is not sufficient to classify the different brain states associated with different cognitive activity. Brain state classification is a challenging task when using standard SVM. Thus, a Kernel-based SVM (k-SVM) has been undertaken in the present study for classification between non-meditation (controlled group) and meditation based EEG. The k-SVM is popularly known as a non-linear classifier. In the present work, a comparative study has been taken up to classify the resting brain state associated with Kriya Yoga meditation practice using SVM and Kernel-SVM (k-SVM). The EEG signals have been captured from ten non-meditators (control group) and 23 meditators group. The results of both SVM and k-SVM have been shown and compared in both the groups. Additionally, the average classification accuracy has been found to be 85.543% for SVM and 90.8259% for k-SVM. The obtained results show that the kernel-based SVM surpassed the conventional SVM in classifying the meditation and non-meditation allied EEG.
Shaw et al. (Thu,) conducted a other in Kriya Yoga meditation practice (n=33). Kernel-based SVM (k-SVM) vs. Conventional SVM was evaluated on Average classification accuracy. Kernel-based SVM (k-SVM) surpassed conventional SVM in classifying meditation and non-meditation allied EEG, achieving an average classification accuracy of 90.8% compared to 85.5%.