Improvements of Vietnamese Hidden Markov Model based speech synthesis

This book aims to present recent advances in Hidden Markov Model (HMM) based speech synthesis with a focus on applications to Vietnamese language. In the last decade, a lot of improvements have been made to HMM-based speech synthesis, making it become the mainstream in speech synthesis research and the popular choice when one desires to develop a text-to-speech (TTS) system for a particular language. Several HMM-based TTS systems for Vietnamese have been developed since 2009. Several improvements have been made to these systems, covering mainly the incorporation of syntactic and prosodic information to enhance the naturalness of the prosody of speech generated by a speaker-dependent model. Although the obtained results were promising, there have been many issues yet to be solved. This book introduces and tackles three problems, which are (i) the modeling of the dynamic features of speech parameters, (ii) the extraction of the fundamental frequency (or F0) parameter in glottalized regions of speech signals, and (iii) the development of a speaker-adaptive HMM-based speech synthesis system for Vietnamese.