Neuro-Theological AI: Machine Learning Models for Simulating Mystical States of Consciousness

Abstract

Understanding consciousness via neural signals is still an open problem in computational neuroscience. In this paper, we present Neuro-Theological AI as an EEG based classification model on Consciousness Emotional State using the EEG Brainwave Dataset: Feeling Emotions. Raw EEG signals will be preprocessed with filtering, artifact elimination, segmentation, and normalization steps. Neurophysiological relevant features of EEG signals will be extracted such as frequency band power (Delta, Theta, Alpha, Beta, Gamma), Spectral Entropy, Functional Connectivity, and Graph- theoretic measures of brain network properties. Then, these features will undergo deep learning with SFL-MobileNetV3 as the deep feature extractor, followed by a Graph Attention Network (GAT) in order to model complex dependencies in the neural process of generating consciousness emotions. The model is evaluated on the commonly used Accuracy, Recall, and F1 Score metrics. Experimental results obtained are high values of 98.2% of accuracy, 97.3% of recall, and 97.5% of F1 score, higher compared to other methods like SVM, DNN, and LSTM.