【PaperReading】DUAL-PATH RNN: EFFICIENT LONG SEQUENCE MODELING FOR TIME-DOMAIN SINGLE-CHANNEL SPEECH

Title：DUAL-PATH RNN: EFFICIENT LONG SEQUENCE MODELING FOR TIME-DOMAIN SINGLE-CHANNEL SPEECH SEPARATION

• What’s main claim? Key idea?

  This paper proposes a simple network called dual-path RNN (DPRNN), that organizes any kinds of RNN layers to model long sequential inputs in a very simple way.

• What’s key limitation?

  Recently, time-domain methods have become popular. The methods in time-domain all rely on effective modeling of extremely long input sequences. This poses an additional challenge as conventional sequential modeling networks, including RNNs and 1-D CNNs, have difficulty on learning such long-term temporal dependency.

• Is there code available? Data?

  no code

  data：WSJ0-2mix and Librispeech

• Is