A Post-Processing Approach to Improve Emotion Recognition Rates

The involvement of emotions in intelligent human-machine interfaces has evolved to a recent field of research. This does not only include the handling of emotions in dialogue management but also the classification of emotions. In this paper we focus on emotion recognition from speech signals and moreover on how to improve recognition results using a post-processing approach. At first we give a short overview on emotion recognition and on corpus which we modified and used in our experiments. Then we describe Fiscus´ ROVER system and the way of how to adapt it to emotion recognition. Finally we present upper and lower bounds for the system´s recognition performance in selected experiments