A survey on the evolution of various voice conversion techniques

Voice conversion (VC) is an area of speech processing that deals with the process of hiding a speaker's identity. This area has been gaining a lot of importance in the recent years. In voice conversion, the voice of the source speaker is taken and modified so that it is made to sound as though it was rendered by the target speaker. Several techniques to voice conversion have been presented and this paper is a survey on those various methods to achieve voice conversion. The techniques trace out from the codebook mapping techniques to Hidden Markovian Model to Gaussian Mixture Model and finally have evolved up to the neural networks and machine learning techniques. These methods focus on various tasks such as speech enhancement, emotion conversion and speaking assistance.