A Pragmatic Solution to an Indian Accented English Speech Synthesizer Using Residual Excited Linear Predictive Coded Voice

This paper elucidates a practical solution to an Indian accented English text to speech synthesizing system. The paper covers the complete procedure to generate the speech signal of the text, in Indian accented voice. The technique described considers the various prosodic features that need to be incorporated into the synthesized speech to make it appear natural and in the way an Indian speaks. The paper describes the complete method for synthesizing the speech including the pre-processing of the text and prosody analysis in the Indian style of speaking. The diphones extracted from an Indian’s voice are coded using Residual Excited Linear Predictive (RELP) coding technique and the resulting formants and residual values of the diphones are used to resynthesize the final output speech for the text to be synthesized. ‘PALSA’ – an Indian accented English speech synthesizer has been successfully implemented using the mentioned technique to produce the Indian accented speech and is described in this paper.