Fig. 4From: Adversarial joint training with self-attention mechanism for robust end-to-end speech recognitionModel architecture of the Transformer. a Encoder. b Decoder [36]Back to article page