Hierarchical Classifiers Approach for Emotions Recognition

ENRIQUE M. ALBORNOZ; DIEGO H. MILONE; HUGO L. RUFINER

Rosario, Santa Fe

Congreso; XIII Reunión de Trabajo en Procesamiento de la Información y Control (RPIC); 2009

The recognition of the emotional states of speaker is a multi-disciplinary research area that has received great interest in the last years. One of the more important goals is to improve the voiced-based human-machine interactions. Recent works on this domain use the prosodic features and the spectrum characteristics of speech signal, with standard classifier methods. However, there is no analysis of what are the causes of the results obtained. Furthermore, for traditional methods the  improvement in performance has also found a limit. In this paper, a study of spectral characteristics of emotional signals is presented. This information is also used in order to group emotions based on their spectral similarities. Hidden Markov Models and Multilayer Perceptron have been evaluated in different configurations with different features, to design a new hierarchical method for emotions classification. Results with the hierarchical method improve up to 6.35% the recognition rate.