Investigation of Using Disentangled and Interpretable Representations for One-shot Cross-lingual Voice Conversion

We study the problem of cross-lingual voice conversion in non-parallel speech corpora and one-shot learning setting. Most prior work require either parallel speech corpora or enough amount of training data from a target speaker. However, we convert an arbitrary sentences of an arbitrary source speaker to target speaker's given only one target speaker training utterance. To achieve this, we formulate the problem as learning disentangled speaker-specific and context-specific representations and follow the idea of [1] which uses Factorized Hierarchical Variational Autoencoder (FHVAE). After train...

Reducing one-to-many problem in Voice Conversion by equalizing the formant locations using dynamic frequency warping

In this study, we investigate a solution to reduce the effect of one-to-many problem in voice conversion. One-to-many problem in VC happens when two very similar speech segments in source speaker have corresponding speech segments in target speaker that are not similar to each other. As a result, the mapper function usually over-smoothes the generated features in order to be similar to both target speech segments. In this study, we propose to equalize the formant location of source-target frame pairs using dynamic frequency warping in order to reduce the complexity. After the conversion, anoth...