Voice conversion based on Genetic Algorithms

Voice conversion is a technique for converting source speaker´s utterance into target speaker´s utterance. Genetic Algorithms (GA) and Gaussian mixture model (GMM) is utilized for performing joint spectrum and pitch conversion between speakers in this paper. The conventional optimization algorithm is orientable and easy to stick in local optimal values. However, Genetic algorithms is probabilistic and holding characteristic of seeking whole optimal values. This capability is very fit of voice conversion. Experimental results indicate that the performance of VC based on genetic algorithms is high-efficiency.