Abstract
In this research, the application of the Corpus-based Singing Voice Synthesis for Mandarin Chinese was proposed. The basic extensions and improvements of the techniques for singing voice synthesis were also proposed. First, the design rules of three corpora for singing voice synthesis for Mandarin Chinese have been presented. Methods of corpora design and preprocessing specifically for the singing voice were also developed. Dynamic programming (similar to Viterbi search) was then applied to select the optimum synthesis units based on the combination of two distance functions. Furthermore, several sound effects, such as echo, vibrato, background music, etc. were implemented to enhance the naturalness of synthesized sounds. Finally, a simple listening test was done for three kinds of synthesis setups to verify the feasibility of this system.