EURASIP Journal on Audio, Speech, and Music Processing

Table 2 Performance in speech denoising

From: Explicit-memory multiresolution adaptive framework for speech and music separation

	BBC			DCASE
Method	PESQ	eSTOI	SDR	PESQ	eSTOI	SDR
BLSTM	1.96	79.77	7.28	1.72	77.92	5.74
CNN	2.08	81.85	7.78	2.15	81.75	6.96
SEGAN	2.02	80.65	6.68	1.99	79.95	6.10
\(L_{1+2-SI}\)	2.34	84.05	9.20	2.51	83.50	7.92
\(TD_{1+2-SI}\)	2.39	84.15	9.68	2.53	83.54	8.52

Best performance is marked bold

Back to article page