Table 2. Speech separation accuracy of ODAN in separating one-, two-, and three-speaker mixtures (WSJ0-mix2 and WSJ0-mix3 datasets).

The ODAN was trained on both the WSJ0-mix2 and WSJ0-mix3 datasets and used in all cases.

Number of speakers	Causal	SI-SNRi (dB)	SDRi (dB)	PESQ	ESTOI
3	Yes	7.0	7.5	2.08	0.56
2	Yes	8.9	9.3	2.63	0.76
1	Yes	SI-SNR (dB) 24.4	SDR (dB) 25.0	4.14	0.98