Speech Enhancement On Demand

评估指标

PESQ (wb)

Para. (M)

评测结果

各个模型在此基准测试上的表现结果

模型名称	PESQ (wb)	Para. (M)	Paper Title	Repository
D2Former	3.43	0.86	Monaural Speech Enhancement with Complex Convolutional Block Attention Module and Joint Time Frequency Losses	-
DeepFilterNet3	3.17	-	DeepFilterNet: Perceptually Motivated Real-Time Speech Enhancement	-
Centaurus (0.51M)	3.25	-	Let SSMs be ConvNets: State-space Modeling with Optimal Tensor Contractions	-
PCS	3.35	-	Perceptual Contrast Stretching on Target Feature for Speech Enhancement	-
MetricGAN-OKD	3.24	1.89	MetricGAN-OKD: Multi-Metric Optimization of MetricGAN via Online Knowledge Distillation for Speech Enhancement
SGMSE+	3.11	-	An Analysis of the Variance of Diffusion-based Speech Enhancement	-
RDL-Net 1.87M (Deep Xi - MMSE-LSA)	2.93	-	Deep Residual-Dense Lattice Network for Speech Enhancement	-
PESQetarian	3.82	30	The PESQetarian: On the Relevance of Goodhart's Law for Speech Enhancement	-
ZipEnhancer (S, lamba_6 = 0.2)	3.61	2.04	-	-
MANNER	3.21	-	MANNER: Multi-view Attention Network for Noise Erasure	-
Dense-TSNet	3.05	0.014	Dense-TSNet: Dense Connected Two-Stage Structure for Ultra-Lightweight Speech Enhancement	-
SEMamba (+PCS)	3.69	2.25	An Investigation of Incorporating Mamba for Speech Enhancement	-
ROSE	3.01	36.98	ROSE: A Recognition-Oriented Speech Enhancement Framework in Air Traffic Control Using Multi-Objective Learning	-
xLSTM-SENet2	3.53	2.27	xLSTM-SENet: xLSTM for Single-Channel Speech Enhancement	-
SCP-CMGAN	3.52	-	SCP-GAN: Self-Correcting Discriminator Optimization for Training Consistency Preserving Metric GAN on Speech Enhancement Tasks	-
aTENNuate	3.27	-	aTENNuate: Optimized Real-time Speech Enhancement with Deep SSMs on Raw Audio	-
SGMSE+ (Diffusion Model)	2.93	-	Speech Enhancement and Dereverberation with Diffusion-based Generative Models	-
real-time-GRU	2.82	-	A Modulation-Domain Loss for Neural-Network-based Real-time Speech Enhancement	-
ZipEnhancer (S, lamba_6 = 0)	3.63	2.04	-	-
MANNER-S + MV-AT (8.1GF)	3.12	1.38	Multi-View Attention Transfer for Efficient Speech Enhancement	-

0 of 40 row(s) selected.