Emotion Analysis Using Spiking Neural Networks

Abstract

Emotion analysis is a prominent research area gaining popularity as more researchers are trying to solve questions in fields that require intensive study of emotions like psychology, human-computer interactions, and affective computing. Music can be used as a tool to identify emotions. NeuCube is a software that considers spatio-temporal data as input and analyzes brain patterns using various encoding, mapping, supervised and unsupervised learning, modelling, network analysis, classification, and optimization techniques. This study used the MUSIN-G dataset to classify the positive emotion 'enjoyment' in healthy participants, employing spiking neural networks with the deSNN classifier and k-fold cross-validation. An accuracy of 83.33% for enjoyment was achieved, with the study identifying stronger connections in brain regions such as AF3, P8, and FC6. Additionally, higher activation levels were observed in the frontal regions, specifically AF3, AF4, and F3. These findings highlight that frontal brain regions process positive emotions.