Fig. 3From: Lightweight target speaker separation network based on joint trainingStructure of the lightweight speaker separation module based on LSTMBack to article page