Fig. 4From: Multimodal voice conversion based on non-negative matrix factorizationAudio wavesBack to article page