Speech enhancement has greatly benefited from deep learning. Currently, the best performing deep architectures use long short-term memory (LSTM) recurrent neural networks (RNNs) to model short and ...