Emotional Prosodic Model Evaluation for Greek Expressive Text-to-Speech Synthesis
- Cite this paper as:
- Tsonos D., Stavropoulou P., Kouroupetroglou G., Deligiorgi D., Papatheodorou N. (2014) Emotional Prosodic Model Evaluation for Greek Expressive Text-to-Speech Synthesis. In: Stephanidis C., Antona M. (eds) Universal Access in Human-Computer Interaction. Universal Access to Information and Knowledge. UAHCI 2014. Lecture Notes in Computer Science, vol 8514. Springer, Cham
In this study we introduce a novel experimental approach towards the evaluation of emotional prosodic models in Expressive Speech Synthesis. It is based on the dimensional emotion expressivity and adopts the Self-Assessment Manikin Test. We applied this experimental approach to evaluate an emotional prosodic model for Greek expressive Text-to-Speech synthesis. We used two pseudo-sentences for each of the Greek and English HMM-based synthetic voices, implemented in the MARY TtS platform. Fifteen native Greek participants were asked to assess eleven emotional states for each sentence. The results show that the “Arousal” dimension is perceived as intended, followed by the “Pleasure” and “Dominance” dimensions’ ratings. These preliminary findings are consistent with the results in previous studies.
KeywordsExpressive Speech Synthesis prosody evaluation Text-to-Speech emotional state
Unable to display preview. Download preview PDF.