Skip to main content
. Author manuscript; available in PMC: 2017 Apr 1.
Published in final edited form as: Speech Commun. 2016 Apr 1;78:1–10. doi: 10.1016/j.specom.2015.12.006

Figure 9.

Figure 9

Mask comparisons. The top shows a ratio mask obtained from training on original noises, the middle shows a mask obtained from training on frequency perturbed noise, and the bottom shows the IRM.