Fig. 2From: Sub-convolutional U-Net with transformer attention network for end-to-end single-channel speech enhancementArchitecture of subconvolutional encoder blocksBack to article page