Indonesian Text-To-Speech system using syllable concatenation: Speech optimization

In the Indonesian language (bahasa), a strong correlation between sounds and letterscan be found, due to the fact that every single letter in a syllable is strongly pronounced. This `feature´ enables the creation of a syllable sound library with a one to one mapping to its written form. A previous investigation^[2] suggest that a total of 1881 syllable sound library is needed to build an Indonesian Text to Speech System (TTS), using the syllable concatenation approach.This paper discuss the optimization of the Indonesian Text-To-Speech (TTS) that has been implemented earlier. The optimization is focused on the syllable sound data base quality and in the integration of syllables during the playback process. Three major improvements is explored here, syllable sound database pre-processing, syllabification method and playback algorithms to reach human reading speed.