A text to speech synthesizer using acoustic unit based concatenation for any Indian language of Devanagari script

This paper presents the prosody improvement in acoustic unit based concatenation of any language of Devanagari script. In Text to Speech synthesis system syllable-based synthesis is quite intelligible as compared to di-phone based systems. The system so developed accept the written text of any language of Devanagari script via MS word utility through MATLAB which then converted into Romanized script under text analysis. Tokenization is performed on Romanized script, for the identification of respective phoneme, syllable and diphthong acoustic units. Using the prepared database of acoustic units, these speech segments undergo concatenation using modified TD-PSOLA technique only at the disjoint points. The proposed work is performed on selecting units based on acoustic framework. The efficient use of TDPSOLA technique synthesizes the speech signal which bears the high degree of accuracy and intelligibility to understand and hear clearly like human speech.