Neural network approach for classification of human emotions from EEG signal

Abstract

Emotions play an important role in human cognition, perception, decision-making, and interaction. In this paper, Neural Network (NN) based system for human emotions classification by extracting features from Electroencephalogram (EEG) signal is proposed. EEG data for the classification of emotions is obtained from the DEAP database. Extracted more than 30 features from EEG and they are used for the emotion classification. Totally, 33 varieties of features are extracted from EEG data. However, there are reports on voice-based, facial-image-based study of expressions to recognize their emotions. However, emotion identification using both methods can be biased as they can be faked. In order to overcome this difficulty, many researchers analyze brain physiological signals to represent the changing patterns during emotional fluctuations. Neural networks have widely been used in emotion classification. Reported here is the classification with the backpropagation artificial neural network. Experimental results have shown an average accuracy above 94.45% is achieved for all the subjects and regions combined.