Towards Real-Time Single-Channel Singing-Voice Separation with Pruned Multi-Scaled Densenets | IEEE Conference Publication | IEEE Xplore