An information-theoretical method for emotion classification

Abstract

Identifying the emotion that someone is feeling will allow to improve the experience of the person interaction with environments, devices, and contents. Our body responds to events around us, by emotional responses, reflected in cognitive, behavioral and physiological dimensions. In the present work, we target the electrocardiogram (ECG) response as a mean to express emotions. Its processing is performed using information-theoretical measures, allowing true exploratory data mining. Participants recruited for the experiment watched three video sets in three different days, with a different emotion being induced in each day: fear, happiness, and neutral condition. The method is divided in: (1) conversion of the real-valued ECG record into a symbolic time-series; (2) relative compression of the symbolic representation of the ECG, using the symbolic ECG records stored in the database as a reference; (3) identification of the ECG record class, using a 1-NN (nearest neighbor) classifier. An accuracy of 90% was obtained. A posteriori analysis of the false negative results indicated that there was a relation between the relative dissimilarity measure and the self-reported emotions.