EURASIP Journal on Audio, Speech, and Music Processing

Table 9 Average SDR scores of compared methods for noisy and enhanced speech under various SNR conditions

From: Supervised Attention Multi-Scale Temporal Convolutional Network for monaural speech enhancement

	No RIR							RIR
SNR	− 5 dB	0 dB	5 dB	10 dB	15 dB	20 dB	\(\Delta\)Avg.	− 5 dB	0 dB	5 dB	10 dB	15 dB	20 dB	\(\Delta\)Avg.
Unprocessed	3.79	8.84	12.84	17.84	22.84	27.84	-	3.73	8.77	12.77	17.77	22.77	27.77	-
CRN	11.24	14.77	17.49	20.90	24.24	27.43	3.68	10.88	14.36	17.09	20.52	23.97	27.25	3.41
MSTCN	9.79	13.22	15.74	18.51	20.71	22.23	1.03	9.44	12.88	15.41	18.26	20.60	22.27	0.88
LSTM-IRM	11.69	15.33	18.20	21.83	25.47	29.18	4.61	11.47	15.03	17.90	21.58	25.45	29.45	4.55
GCRN	12.71	16.06	18.64	21.79	24.81	27.55	4.59	11.91	15.27	17.89	21.09	24.22	27.01	3.97
GaGNet	12.27	15.62	18.30	21.55	24.79	28.09	4.43	12.15	15.49	18.13	21.54	25.02	28.68	4.57
Conv-TasNet	13.91	17.10	19.55	22.61	25.73	28.99	5.65	13.09	16.33	18.89	22.18	25.65	29.32	5.31
DCCRN	13.65	17.22	20.04	23.44	26.87	30.45	6.28	12.74	16.22	18.96	22.43	25.93	29.68	5.39
DPCRN	13.33	16.78	19.45	22.76	26.01	29.11	5.57	12.92	16.32	19.02	22.40	25.80	29.25	5.35
SA-MSTCN\(^{1}\)	13.34	16.68	19.33	22.64	26.01	29.51	5.58	13.45	16.74	19.24	22.69	26.32	30.08	5.82
SA-MSTCN\(^{2}\)	13.66	17.05	19.63	22.91	26.26	29.62	5.85	13.45	16.74	19.47	22.87	26.46	30.15	5.92

Back to article page