EMOTIONAL VOICE CONVERSION WITH WAVELET TRANSFORM USING DUAL SUPERVISED ADVERSARIAL NETWORKS

摘要

This paper proposed an emotional VC methodusing dual supervised adversarial networks (dual-SANs) with MCC and CWT-F0 features. In orderto obtain a better training results, we process 2DMCC features and 2D AS-CWT features, and sep-arately train them with the proposed dual-SANs.The results shows that the dual supervised learninghas a great affect on the dual emotional VC tasks.