Abstract:
Text to Speech (TTS) Synthesizer is an application that converts text to speech. A TTS has applications that make computer systems interactive and help its users especially the visually challenged. Concatenative synthesis technique uses different units of speech such as words, syllables, diaphones and phonemes. Most of the Indian languages are syllabic in nature and thus syllables are best suited as the unit of synthesis over phonemes and diphones. Syllabification is used in Speech Synthesis Systems in producing natural sounding speech and in speech recognizers in detecting words. This paper investigates the use of some candidate speech synthesizers inclusive of Indian and non-Indian languages in the context of high quality TTS Systems using syllable as the unit for synthesis. Analysis indicates that the syllable based approach performs better for Indian languages. Forward and backward approach to generate syllables is implemented for Indian languages written in Devanagari Script like Hindi, Marathi and Konkani for the development of a TTS system. The system is developed for different units of speech such as words, syllables, diphones and phonemes. The test results indicate positive results in terms of naturalness for Konkani language taking into account the implementation of the Konkani phonological rules. Each language has specific phonological rules requiring attention for the development of a TTS system with high naturalness and intelligibility rather than only using syllable as the unit for synthesis. Thus it becomes complex to develop a common generic TTS system for different languages.
