Corpus-based Malay text-to-speech synthesis system

The main problem with current Malay text-to-speech (TTS) synthesis system is the poor quality of the generated speech sound. This poor quality is resulted from the inability of traditional TTS system to provide multiple choices of unit for generating more accurate synthesized speech. Most of the current available Malay TTS systems are utilizing diphone concatenation that only support a single unit for each existing diphone, thus it cannot provide more accurate selection of speech unit for concatenation. This project has implemented a variable length unit selection Malay text to speech system that is capable of providing more natural and accurate unit selection for synthesized speech. This paper proposes a method of combining both linguistic context and feature distance cost for selecting the best match unit. A set of digitized Malay word has been collected from Malay Internet news for Malay word frequency count. 381 sentences have been designed which cover around 70 percent of high frequency words from 10 million digitized word obtained from Malay Internet news. Then a unit selection method has been implemented to provide the capability of selecting a speech unit not only limited to phoneme, diphone or triphone but also a string of phonemes that can be matched directly to the database. A set of listening test namely modify rhythm test (MRT) has been carried out with 35 participants, which represented 86 percent of accuracy.