Semi-supervised Monaural Singing Voice Separation with a Masking Network Trained on Synthetic Mixtures | IEEE Conference Publication | IEEE Xplore