Table 1 Comparison of speech separation accuracy of ODAN with two other methods for separating two-speaker mixtures (WSJ0-mix2 dataset) and three-speaker mixtures (WSJ0-mix3 dataset).

Comparison of speech separation accuracy of ODAN with two other methods for separating two-speaker mixtures (WSJ0-mix2 dataset) and three-speaker mixtures (WSJ0-mix3 dataset).. The separation accuracy of ODAN, which is the causal system, is slightly worse but comparable to the other noncausal methods.

Number of
Speakers
ModelCausalSI-SNRi
(dB)
SDRi
(dB)
PESQESTOI
Two speakersOriginal mixture002.020.56
DAN-LSTM (11)No9.19.52.730.77
uPIT-LSTM (15)Yes7.0
ODANYes9.09.42.700.77
Three speakersOriginal mixture001.660.39
DAN-LSTM (11)No7.07.42.130.56
uPIT-BLSTM (15)No7.4
DPCL++ (50)No7.1
ODANYes6.77.22.030.55