Pitch conversion based on pitch mark mapping

Abstract

Voice conversion is a subject of major research interest these days, due to its numerous applications including dubbing, text to speech synthesis and multimedia. Pitch conversion is an important building block for efficient voice conversion. This paper focuses on the implementation of pitch scaling and pitch conversion. Effective silence removal, voice/unvoiced detection, pitch extraction, pitch marking and PSOLA techniques have been implemented for pitch scaling along with an implementation of pitch mark mapping method for pitch conversion. A graphical user interface has been developed for evaluation and illustration purposes. Evaluation of the pitch scaling and pitch conversion technique was performed on various speech data. Pitch scaling and pitch conversion for both the male and female speakers was implemented and verified with the help of pitch contour plots of the transformed speech.