Unsupervised Sound Separation Using Mixture Invariant Training

In recent years, rapid progress has been made on the problem of single-channel sound separation using supervised training of deep neural networks.