Classification of Three Negative Emotions based on Physiological Signals

Physiological signal is one of the most commonly used emotional cues. In recent emotion classification research, the one of main topics is to recognize human's feeling or emotion using multi-channel physiological signals. In this study, we discuss the comparative results of emotion detection using several classification algorithms, which classify negative emotions (fear, surprise and stress) based on physiological features. Physiological signals, such as skin temperature (SKT), electrodermal activity (EDA), electrocardiogram (ECG), and photoplethysmography (PPG) were recorded while participants were exposed to emotional stimuli. Twenty-eight features were extracted from these signals. For classification of negative emotions, four machine learning algorithms, namely, Linear Discriminant Analysis (LDA), Classification And Regression Tree (CART), Self Organizing Map (SOMs), and Naive Bayes were used. The 70% of the whole datasets were selected randomly for training and the remaining patterns are used for testing purposes. Testing accuracy by using the 30% datasets ranged from 32.4% to 46.9% and, consequently the selected physiological features didn't contribute to classify the three negative emotions. In the further work, we intend to improve emotion recognition accuracy by applying the selected significant features, such as NSCR, SCR, SKT, and FFTap_HF.