Fig. 7From: Progressive loss functions for speech enhancement with deep neural networksMSE between clean reference and reconstruction output at each block on the REVERB-Dev set. The dark line shows the mean and the shaded area around shows the standard deviationBack to article page