Singing voice synthesis for Indian Classical Raga system

In this paper, a speech-to-singing voice synthesis system is presented, built using MATLAB platform and PRAAT speech processing system, oriented to produce singing voice for Indian Classical Raga framework. The system modifies the various acoustic features essential for singing voice, like the duration and pitch (fO) contour of the input neutral speech to convert it into singing voice, using the given musical information, viz, lyrics and notes arranged into matras (beats). The paper presents pitch contour shaping methods to improve the naturalness of the synthesized singing voice. A fixed parameter second order filter and an adaptive recursive filter have been used to generate fundamental frequency contour. Voice synthesis performances have been compared.