Learning a cross-domain embedding space of vocal and mixed audio with a structure-preserving triplet loss

Keunhyoung Luke Kim, Jongpil Lee, Sangeun Kum, Juhan Nam