A Hierarchical Classification Scheme for Efficient Speech Emotion Recognition

The current study focuses on speech emotion recognition based on a hierarchical classification scheme. The study aims at overcoming the problem of low accuracy in the case of a large number of emotions that are considered in a specific task. In the proposed method, the emotions are classified based on the valence-arousal 2-dimensional map, and models are trained for each group. In a second pass, with-in group recognition is performed for the group selected in the previous stage.