Zero-mean Convolutional Network with Data Augmentation for Sound Level Invariant Singing Voice Separation | IEEE Conference Publication | IEEE Xplore