Fig. 4From: Lightweight target speaker separation network based on joint trainingExperimental results on Librispeech test setBack to article page