Fig. 2From: A multi-task learning speech synthesis optimization method based on CWT: a case study of Tacotron2Deep learning acoustic models with multi-task learningBack to article page