Phonetic posteriorgrams for many-to-one voice conversion without parallel data training | IEEE Conference Publication | IEEE Xplore