Table 5. Unmatched test scores in five noise sources at all SNRs.

The results are averaged over all testing utterances.

Noise	Algorithm	STOI (in%)	PESQ	SDR
Airport Noise	Noisy (UnP)	71.6	1.83	0.14
	DNN-IRM	82.6	2.27	6.30
	LSTM-IRM	85.0	2.37	6.84
	LSTM-KF	85.9	2.40	6.96
	FDNN-KF	83.4	2.31	6.41
	DeepXi-KF	86.7	2.45	7.21
	DeepResGRU-KF	86.9	2.46	7.31
	Proposed	87.6	2.48	7.39
Babble Noise	Noisy (UnP)	65.9	1.72	0.19
	DNN-IRM	76.7	2.12	6.13
	LSTM-IRM	78.5	2.23	6.31
	LSTM-KF	79.1	2.28	6.35
	FDNN-KF	77.2	2.15	6.18
	DeepXi-KF	80.2	2.30	6.45
	DeepResGRU-KF	80.8	2.32	6.51
	Proposed	81.0	2.35	6.62
Car Noise	Noisy (UnP)	67.5	1.65	0.13
	DNN-IRM	79.6	2.17	6.38
	LSTM-IRM	82.8	2.33	6.92
	LSTM-KF	83.1	2.40	7.01
	FDNN-KF	80.1	2.21	6.45
	DeepXi-KF	84.2	2.45	7.12
	DeepResGRU-KF	84.8	2.48	7.28
	Proposed	86.0	2.49	7.45
Factory Noise	Noisy (UnP)	66.1	1.61	0.18
	DNN-IRM	75.7	2.08	5.92
	LSTM-IRM	78.6	2.24	6.39
	LSTM-KF	78.8	2.29	6.45
	FDNN-KF	76.0	2.10	8.98
	DeepXi-KF	79.5	2.35	6.57
	DeepResGRU-KF	80.2	2.38	6.71
	Proposed	81.3	2.39	6.87
Street Noise	Noisy (UnP)	66.8	1.73	0.22
	DNN-IRM	77.8	2.18	6.06
	LSTM-IRM	81.5	2.31	6.34
	LSTM-KF	81.9	2.35	6.40
	FDNN-KF	78.0	2.21	6.10
	DeepXi-KF	82.6	2.38	6.45
	DeepResGRU-KF	84.7	2.40	6.58
	Proposed	85.1	2.42	6.63