ext to Speech synthesizer for Dzongkha

Abstract

A high-quality speech synthesizer should be intelligent and produce natural speech. The quality of speech generated by a text-to-speech synthesizer also depends on the amount of data used for training. This paper presents the development of Dzongkha TTS (Text-to-Speech) system using open-source toolkit Hidden Markov Model Toolkit (HTK) and proposes a method to increase the speech database for quality output. Every word in a language can be broken down into several phonemes, which are combinations of phonemes that generate words. Therefore, we suggest developing a corpus through phoneme concatenation, which can increase the database for training TTS systems.