Target Speaker Extraction by Directly Exploiting Contextual Information in the Time-Frequency Domain | Synapse