R.-H.Wang, “Articulatory control of HMM-based parametric speech synthesis driven by phonetic knowledge

Abstract—In previous work we proposed amethod to control the characteristics of synthetic speech flexibly by integrating articu-latory features into a hidden Markov model (HMM) based para-metric speech synthesizer. In this method, a unified acoustic-ar-ticulatory model is trained, and context-dependent linear trans-forms are used to model the dependency between the two feature streams. In this paper, we go significantly further and propose a feature-space-switched multiple regression HMM to improve the performance of articulatory control. A multiple regression HMM (MRHMM) is adopted to model the distribution of acoustic fea-tures, with articulatory features used as exogenous “explanatory” variables. A separate Gaussian mixture model (GMM) is intro-duced tomodel the articulatory space, and articulatory-to-acoustic regression matrices are trained for each component of this GMM, instead of for the context-dependent states in the HMM. Further