Emotion analysis from speech using temporal contextual trajectories

We propose a framework for speech emotion analysis that maps sequential acoustic features into descriptors that integrate the temporal ordering of the original features and the discrimination between emotions. Using the topology preserving property of Self Organizing Maps and the continuous nature of speech, the proposed framework maps a speech utterance into a temporally ordered trajectory that reflects the emotion activity over time in the underlying utterance. Using a standard emotional database, we show that the proposed framework is efficient for analysis, visualization and classification of emotions.