Skip to main content
. 2017 Jun 23;141(6):4705–4714. doi: 10.1121/1.4986931

FIG. 6.

FIG. 6.

(Color online) Visualization of the estimated masks by the DNN (top), LSTM (middle), and the IRM (bottom). The mixture is created by mixing an unseen male speaker with the unseen babble noise at −5 dB SNR.