The Role of Empirical Mode Decomposition on Emotion Classification Using Stimulated EEG Signals

Abstract

An efficient scheme of emotion recognition using EEG signals is an initiation to our quest for developing emotionally intelligent systems and devices, in order to enhance the performance quality of the same. Classification of emotions, both euphoric and negative, using stimulated EEG signals acquired from subjects whose different emotional states were elicited using audio-visual stimuli. The underlying strategy involved the extraction of Power spectral density(PSD) and empirical mode decomposition (EMD) features from the raw EEG data and their classification using linear discriminant analysis (LDA) and linear support vector machine (SVM) thereby classifying the emotions into their respective emotion classes: neutral, happy and sad, with an average classification accuracy of 76.46%,where the neutral state has been classified most efficiently, with an average classification accuracy of 80.86%. The classification accuracy increases with EMD features with reduction in time and computational complexity. LDA is found to perform better than LSVM with EMD features.