Robust estimation of multiple-regression HMM parameters for dimension-based expressive dialogue speech synthesis

This paper describes spontaneous dialogue speech synthe- sis based on multiple-regression hidden semi-Markov model (MRHSMM), which enables users to specify paralinguistic in- formation of synthesized speech with a dimensional representa- Tion. Paralinguistic aspects of synthesized speech are controlled by multiple regression models whose explanatory variables are abstract dimensions such as pleasant-unpleasant and aroused- sleepy. For robust estimation of the regression matrices of the MRHSMM with unbalanced spontaneous dialogue speech sam- ples, the re-estimation formulae were derived in the framework of the maximum a posteriori (MAP) estimation. The result of a perceptual experiment confirmed that the naturalness of synthe- sized speech was improved by applying the MAP estimation for regression matrices. In addition a high correlation (R ≃ 0:7) wasobserved between given and perceived paralinguistic infor- mation, which implies that the proposed method could success- fully reflect intended paralinguistic messages on the synthesized speech.