Investigating the neurological effects of loving kindness meditation using EEG-based feature extraction and machine learning

Loving-Kindness Meditation (LKM) is a practice developed by the Buddhist community. This method involves a scientific approach to projecting well wishes and compassionate thoughts, first to oneself and then to others. This paper examines the neurological effects of LKM by analysing existing electroencephalographic (EEG) signals that categorizes four distinct mental states associated with LKM, including self-directed, other-directed, and baseline conditions before and after the LKM. The dataset used consists of measurements taken before meditation (PreResting), after meditation (PostResting), LKMSelf, and LKMOther states, including 127-channel EEG recordings for 32 subjects, out of which 15 have multiple sessions. Standard preprocessing of EEG signals includes bandpass filtering, and artifact removal is conducted using independent component analysis (ICA). Common Spatial Patterns (CSP), Riemannian geometry, and frequency band powers are used for feature extraction. Linear Discriminant Analysis (LDA) and other machine learning (ML) techniques are employed to classify different meditation states. Based on the findings, it has been demonstrated that EEG features, both in the spatial and frequency domains, can effectively classify meditative and non-meditative states across multiple sessions. The high classification accuracies indicate that this system generalizes across different states of meditation and non-meditation.