Interspeech'2005 - Eurospeech

Lisbon, Portugal
September 4-8, 2005

Performance Evaluation of Style Adaptation for Hidden Semi-Markov Model Based Speech Synthesis

Makoto Tachibana, Junichi Yamagishi, Takashi Masuko, Takao Kobayashi

Tokyo Institute of Technology, Japan

This paper describes a style adaptation technique using hidden semi-Markov model (HSMM) based maximum likelihood linear regression (MLLR). The HSMM-based MLLR technique can estimate regression matrices for affine transform of mean vectors of output and state duration distributions which maximize likelihood of adaptation data using EM algorithm. In this study, we apply this adaptation technique to style adaptation in HSMM-based speech synthesis. From the results of several subjective tests, we show that the HSMM-based MLLR technique can perform style adaptation with maintaining naturalness of the synthetic speech compared with the conventional HMM-based MLLR technique.

Full Paper

Bibliographic reference.  Tachibana, Makoto / Yamagishi, Junichi / Masuko, Takashi / Kobayashi, Takao (2005): "Performance evaluation of style adaptation for hidden semi-Markov model based speech synthesis", In INTERSPEECH-2005, 2805-2808.