Electroencephalogram (Eeg)-Conformer For Emotion Recognition And Theory-Guided Music Prompt Generation: A Prototype Brain-Computer Music Interface

Abstract

There is a robust body of research demonstrating that emotional states profoundly affect both physical and mental health. Acknowledging this, researchers have sought to examine methods for detecting an individual’s emotional state and thus promoting their emotional and psychological well-being. In recent years, scholars working in this area have turned to computer science and machine learning to serve these objectives. In this same vein, the current study examines electroencephalogram (EEG) signals as a means for emotion recognition and for facilitating brain-computer interfaces. To assess the potential of EEG signals for such purposes, this article reviews the principles underlying methods for measuring EEG signals, including frequency bands and electrode distributions, to examine how EEG data can be used for emotion recognition. To inquire into the patterns inherent in brainwave recordings and determine whether a person was experiencing positive, negative, or neutral emotions, this study used a specialized artificial intelligence (AI) model called an EEG-Conformer. Once an emotional state has been recognized, the system pipeline translates the classification result into a text prompt, which is then passed to an AI music-generation model that produces music aligned with the detected state. Testing on a widely used EEG emotion dataset (the SEED dataset) showed that the model was over 95% accurate in identifying emotional states, and visual analyses further confirmed that the model was able to ascertain meaningful patterns in the brain activity associated with different emotions. In this way, findings from this study point to how AI and neuroscience can be integrated in such a way that it supports human health and well-being.