Abstract:
Emotion recognition has gained major importance in recent years, with
applications in human-computer interfaces, affective computing, and numerous
medical applications. To capture and analyze the emotional states, several modalities
are used, where one of the most dominant is Electroencephalography (EEG).
Facilitated by the advancements in EEG acquisition technologies, as well as in the
Artificial intelligence field, Emotion Recognition with EEG data has attracted many
researchers. This work aims to implement a subject-independent model that utilizes
EEG to perform Emotion Recognition on DEAP and DREAMER datasets. It attempts
to find the right combination of processing methods, feature extraction, feature
selection and classifier that generalize well on unseen data without having excessive
computational costs. In this thesis several Machine Learning models are implemented,
along with a one-dimensional CNN model which succeeds in providing a reliable
performance for the task of Emotion Recognition with EEG.
