Characterization and Incorporation of Emotions in Speech

Emotional speech plays an important role for conveying the desired message. Emotions are manifested in speech signal at all the levels, in particular they are significant at suprasegmental level (i.e., prosodic level). In this paper four emotions are characterized (anger, compassion, happy and neutral) using the prosodic features such as duration, intonation (variation of pitch), and energy. The analysis is performed on SUSE (speech under simulated emotion) database. The derived knowledge with respect to each emotion in terms of the prosodic features is used for speech transformation from one emotion to other. In this paper an attempt is made for transforming speech at gross level from neutral to other emotions. The transformed speech after incorporating the emotions is evaluated by using the listening tests