Fig. 2From: Multimodal voice conversion based on non-negative matrix factorizationMultimodal dictionary constructionBack to article page