Publication

Abstract : One of the most commonly used non-invasive techniques for emotion recognition is the electroencephalogram (EEG). EEG can be used to record electrical activity in the brain and cannot be voluntarily fabricated. The electroencephalogram (EEG) is a physiological indicator that shows how electrical activities of brain cells cluster across the human cerebral cortex. Research works that demonstrate how the most complete characteristics of EEG, such as Power Spectral Density (PSD) can be used to classify basic emotions. This paper proposes an efficient method for predicting human emotions using spatial-spectral aspects of EEG and Convolutional Neural Network (CNN). To create a 3D map, spectral features such as PSD and Differential Entropy (DE) are extracted. The 3D brain map is used as an input to a CNN model for classifying emotions into valence and arousal by producing accuracy of 89.38% and 90.12% respectively.