Analyzing brain data for robust emotion recognition via conceptual decomposition based on autoencoders

Recognizing emotions using biological brain signals requires accurate and efficient signal processing and feature extraction methods. The ability to accurately recognize and understand emotions is crucial in various domains, including neuroscience, medicine, psychology, technology and beyond. From a psychological and neuroscientific point of view, emotion recognition from brain signals provides researchers with a powerful tool to study the intricate interplay between brain activity and emotions. The aim of identifying specific brain regions is important to study the basis of emotional processing and this knowledge contributes to a better understanding of emotional disorders, social interactions, cognitive processes. The problem addressed in this thesis work is how to use brain signals for emotion recognition in a profitable and efficient way. Recent studies have shown how these brain regions have different behaviors in emotion recognition task. It has been demonstrated, in fact, that some brain regions are more important than others in this type of task. For this reason,the goal of this thesis is to address this problem by developing an architecture based on 'conceptual decomposition' that exploits two different approches, like Concept Importance and TCAV