Speech Emotion Recognition Using Hierarchical Approach Based On Bhattacharyya Distance

Abstract

In human-human interaction emotions play an essential role in conveying the information apart from the verbal communication. It is quite challenging for machines to recognize human emotions and respond accordingly. Most research in emotion recognition has focused on using the flat approach in which emotions are classified in a single step using a single best set of features. This paper presents a hierarchical approach based on Bhattacharyya distance for human emotion recognition from speech. The basic aim is to improve the emotion classification performance. The analysis is performed using the Surrey Audio-Visual Expressed Emotion (SAVEE) database, while various attribute selectors and classification techniques are implemented to obtain the best results. The experimental results showed better performance for the proposed hierarchical approach as compared to flat approach and other state-of-the-art techniques. The best classification accuracy of 78.12% is achieved for the flat approach, while the best performance of 91.66% is obtained for the hierarchical approach using seven emotions of the SAVEE database.