Annotating conversational speech for corpus-based dialogue speech synthesizer — A first step

This paper describes an HMM-based speech synthesis that allows dimensional description of emotion as inputs. A spontaneous dialogue speech corpus that was designed for studying paralinguistic phenomena in expressive social interactions was used to train the models, utilizing its emotional state description as additional contextual factors. In the perceptual experiment, a very high correlation was observed (R ≃ 0.8) between given pleasantness/arousal values and averaged subjective evaluations, which means that the synthesized utterances could successfully convey specified paralinguistic information.