Voice Tranformation Algorithm for Speech Conversion Applications

Abstract

The basic purpose of the voice Morphing system is to change the characteristics of the source speaker voice in such a way so as it is to be perceived as a target speaker. It is the most important problem these days in the area of voice and speech processing with increasing commercial interest due to applications such as personi cation of text to speech, secured transmission, and speech to speech translations. It is also used in many areas like entertainment, lm, music industries, toys and chat rooms, games, and voice restoration. The process of the voice conversion involves transforming the speaker speci c acoustic cues at di erent levels such as shape of vocal tract, the shape of glottal pulse and the prosodic parameters. In this Project report the basics of speech morphing and the algorithms available in literature are discussed. The physics of speech production which describes how voice is generated in human vocal tract are also discussed. Di erent technique to morph speech signals between source and target speaker are carried out. We considered the source speaker to be a Man or a Woman and the Corresponding target speaker to be a Woman or a man respectively. There are three important parts associated with the voice morphing: LPC based source- lter decomposition, Interpolation of signals between the Source and Target signal, combining everything together. In the rst part of minor project tenure, literature review and LPC based voice morphing technique is considered. Better interpolation techniques and more robust & e ective voice morphing algorithms or frameworks are proposed in the Second part.