Speech Enhancement Using Joint DNN-NMF Model Learned with Multi-Objective Frequency Differential Spectrum Loss Function

Table 2

The noise classification (NC) performance for the proposed Jnt-DNN-NMF-MOFD model.

	Matched noises			Mismatched noises
	PESQ	STOI	SNRfw	PESQ	STOI	SNRfw
Noisy	2.10	0.70	9.30	2.08	0.74	6.11
Without NC	3.10	0.87	13.96	2.35	0.78	8.73
With NC	3.20	0.88	15.27	2.42	0.79	9.51