Analysis of EEG Signals in the DEAP Dataset for Emotion Recognition using Deep Learning Algortihms

Emotions are the behavioral responses representing mental state of a person. It is crucial to recognize the emotions of a person for human-computer interaction, to understand and respond to one's mental health. EEG signal provides a clear-sighted analysis of emotional state. It is a challenging task to recognize the patterns of multi channel EEG signals for emotion recognition using traditional Machine Learning approach. So, the Deep Learning algorithms are preferred more as they are successful in learning features and patterns to classify the the data. In this paper we have proposed Deep Learning models, CNN and LSTM Neural Networks to recognize the emotional states: Arousal, Valence and Dominance for the subjects 01-22 whose frontal videos were recorded and 23-32 whose frontal videos were non recorded in the DEAP dataset. CNN and LSTM Neural Networks are built with significant layers and the model is trained with required number of epochs and batch size after a number of trials to improve the accuracy of the emotion recognition. Finally, performance of the proposed models is evaluated with different evaluation metrics and the accuracy of both the Deep Learning models are compared for the subjects 01-22 and 23-32.