Fig. 3From: Supervised Attention Multi-Scale Temporal Convolutional Network for monaural speech enhancementPrinciples of supervised attention in the U\(^2\)-LSTMBack to article page