High accurate model-integration-based voice conversion using dynamic features and model structure optimization

This paper combines a parameter generation algorithm and a model optimization approach with the model-integration-based voice conversion (MIVC). We have proposed probabilistic integration of a joint density model and a speaker model to mitigate a requirement of the parallel corpus in voice conversion (VC) based on Gaussian Mixture Model (GMM). As well as the other VC methods, MIVC also suffers from the problems; the degradation of the perceptual quality caused by the discontinuity through the parameter trajectory, and the dif culty to optimize the model structure. To solve the problems, this paper proposes a parameter generation algorithm constrained by dynamic features for the rst problem and an information criterion including mutual in uences between the joint density model and the speaker model for the second problem. Experimental results show that the rst approach improved the performance of VC and the second approach appropriately predicted the optimal number of mixtures of the s...

• Type: Journal
• Year: 2011
• Where: ICASSP
• Authors: Daisuke Saito, Shinji Watanabe, Atsushi Nakamura, Nobuaki Minematsu
• Keywords: ICASSP 2011, Joint Density Model, Model, Signal Processing, Speaker Model