Word-based Myanmar Text-to-Speech with CLUSTERGEN

Abstract

In Myanmar language, word information is important for the naturalness of Myanmar speech because pronunciation changes depend on word information. In this paper, Myanmar speech synthesis with word information has been implemented by using CLUSTERGEN. The comparison of the naturalness of Myanmar synthesized speech by using HMM-based model with syllable information and CLUSTERGEN model with word information has been conducted based on Comparison Mean Opinion Score (CMOS). The results show that word information can promote the naturalness of the synthesized speech for Myanmar language.