Emotion classification is essential for understanding human interactions and hence is an important module in Human-Computer Interaction (HCI) systems. Well-performed emotion classification systems have potential to integrate into the HCI systems to provide additional user state details. This paper presents an emotion classification system that employs Emotional Dissimilarity (ED) measure. Instead of measuring ED in Single Dimensional (SD) space, we propose an approach to measure ED in Multi-Dimensional (MD) space. Proposed approach measures ED between emotions using GMM-SVM kernel with Bhattacharyya based GMM distance. In addition, we also formulate ED measure using GMM-SVM kernel with Kullback Leibler (KL) based GMM distance. We observe the effectiveness of employing ED measure in MD space over that in SD space using different kernels. Experiments were conducted using SVM classifier to classify emotions of anger, happiness, neutral and sadness. We achieve average accuracy of 81.25% for speaker independent emotion classification.
Bibliographic reference. Nwe, Tin Lay / Nguyen, Trung Hieu / Limbu, Dilip Kumar (2013): "Bhattacharyya distance based emotional dissimilarity measure in multi-dimensional space for emotion classification", In INTERSPEECH-2013, 1355-1359.