A novel classification of meditation techniques via optimised chi-squared 1D-CNN method based on complexity, continuity and connectivity features
Abhishek Jain, Rohit Singh Raja, Manoj Kumar, Pawan Kumar Verma · Connection Science · 2025
The intricate world of human–computer interaction deeply explores how people gain knowledge and blend technology into daily life. Electroencephalography (EEG) is one of several methods for measuring brain activity, it is non-invasive, portable, inexpensive and time-sensitive. Research shows a strong link between meditation and changes in EEG patterns, spanning various techniques. With machine learning playing a major role, EEG datasets have made comprehensive study possible. This paper investigates the efficacy of 1D-CNN (One-Dimensional Convolutional Neural Network) classification, using complexity, continuity and connectivity features. It remarkably outperforms and achieves 60% training accuracy, showcasing model robustness in meditation classification. This novel methodology enables to differentiates neural oscillations in type of meditator and control. Prior research used power spectrum density, entropy, and connectivity for meditation distinctions. EEG data from practitioners of Himalayan Yoga (HYT), Isha Shoonya (SYN) and Vipassana (VIP) as well as untrained controls (CTR) are examined in this research. Employing chi-square, CNN and hyperparameter models, outcomes reveal distinctive cognitive aspects among meditation styles, allowing effective differentiation.