An affective BCI system with music in an immersive environment

Abstract

This study investigates the effectiveness of generated music in a Virtual reality (VR) immersive environment in inducing different emotional arousal states in the context of alleviating mood disorders. It details the collection and labelling of EEG data from 20 participants which is evaluated using two state-of-the-art EEG emotion classification models – TSception and EEGNet. This report outlines the end-to-end implementation of a novel data recording system combining Python-based music generation, VR development with Unity, and EEG data streaming and labelling. It also highlights the overall system design considerations and experiment protocols administered. Given a fixed high valence value, the labelled EEG data recorded was used for offline model training to classify three emotional arousal states – low, high, and neutral. High classification accuracies were reported for Low-High arousal classification - 81.57% for TSception and 83.45% for EEGNet. Therefore, it can be concluded that the EEG data collected contained distinctive emotional states. This system, combining the effect of VR and music, is effective in inducing emotional arousal states and can be explored further in clinical trials as a potential tool for emotion modulation in alleviating mood disorders.