Text to Personalized Speech Synthesis: Motivation. State of the Art, and Proposed Techniques

摘要

In today's fast paced world we depend on our voices to a great degree to convey our personal ideas and thoughts. The importance of our voice is sometimes taken for granted as we use it every single day of our lives. The loss of the natural voice can have a deep psychological impact on a person. In this paper we use Linear Predictive Coding as one of the techniques to extract the parameters that characterize a person's speech. The main objective behind personalized text to speech is to use these extracted parameters to reconstruct a person's speech in the event of loss of natural voice. Certain email applications is another objective. Some simulations are performed to explore the effects of mixing one speaker's vocal parameters with the other. The simulation results validate the uniqueness of these parameters as a sense of identity for a particular speaker.