An Unit Selection based Hindi Text To Speech Synthesis System Using Syllable as a Basic Unit

Concatenative speech synthesis using phoneme, di-phone and allophone as an elementary unit for Hindi speech synthesis requires significant quality improvement. The naturalness of the state of the art waveform synthesizer is attributed due to the use of syllable as a basic unit. The primary reason for choosing the syllable as a basic unit is that the Indian languages are syllable centered. This work proposes a syllable based speech unit for concatenative speech synthesis considering position of syllable in a word into account i.e the start, middle and end. This is achieved by building a standard syllable (C*V) level speech database consisting of 442 syllables in each position thus accounting for 1326 standard and non-standard words. Further the quality of synthesized speech is enhanced using moving average windowing. The effectiveness of the system is demonstrated by synthesizing natural sounding speech for Hindi, national language of India. An important advantage of this approach leads to reduced prosody mismatch and spectral discontinuity that occurs during syllable concatenation. The results obtained from the proposed system are far superior compared to the traditional unit based Text to Speech (TTS) synthesis system. The most important quality of this system is the improved naturalness in the synthesized speech.