Speech Enhancement on Deep Noise Suppression (DNS) Challenge

Metric: PESQ-WB (higher is better)

LeaderboardDataset

Loading chart...

Results

Hide extra data

Sort:

#	Model↕	PESQ-WB▼	Extra Data	Paper	Date↕	Code
1	ZipEnhancer (M)	3.81	No	-	-	-
2	TF-Locoformer (M)	3.72	No	TF-Locoformer: Transformer with Local Modeling b...	2024-08-06	Code
3	ZipEnhancer (S)	3.69	No	-	-	-
4	MambAttention	3.671	No	MambAttention: Mamba with Multi-Head Attention f...	2025-07-01	Code
5	MP-SENet	3.62	No	Explicit Estimation of Magnitude and Phase Spect...	2023-08-17	Code
6	xLSTM-SENet	3.588	No	MambAttention: Mamba with Multi-Head Attention f...	2025-07-01	Code
7	BSRNN-S + MRSD	3.53	No	High Fidelity Speech Enhancement with Band-split...	2022-12-01	Code
8	BSRNN-16k	3.45	No	High Fidelity Speech Enhancement with Band-split...	2022-12-01	Code
9	MFNET	3.43	No	A Mask Free Neural Network for Monaural Speech E...	2023-06-07	Code
10	BSRNN-S	3.42	No	High Fidelity Speech Enhancement with Band-split...	2022-12-01	Code
11	BSRNN	3.32	No	High Fidelity Speech Enhancement with Band-split...	2022-12-01	Code
12	CleanUNet-2	3.262	No	CleanUNet 2: A Hybrid Speech Denoising Model on ...	2023-09-12	-
13	FRCRN	3.23	No	Monaural Speech Enhancement with Complex Convolu...	2021-02-03	Code
14	FullSubNet+	3.218	No	FullSubNet+: Channel Attention FullSubNet with C...	2022-03-23	Code
15	CleanUNet	3.146	No	Speech Denoising in the Waveform Domain with Sel...	2022-02-15	Code
16	aTENNuate	2.98	No	aTENNuate: Optimized Real-time Speech Enhancemen...	2024-09-05	-
17	Sudo rm -rf (U=32)	2.95	No	RemixIT: Continual self-training of speech enhan...	2022-02-17	Code
18	DCTCRN-P	2.82	No	-	-	-
19	DCTCRN-T	2.82	No	-	-	-
20	DCCRN-E	2.79	No	-	-	-
21	PoCoNet	2.7885	No	PoCoNet: Better Speech Enhancement with Frequenc...	2020-08-11	-
22	FullSubNet	2.777	No	FullSubNet: A Full-Band and Sub-Band Fusion Mode...	2020-10-29	Code
23	DCTCRN-S	2.77	No	-	-	-
24	RNN-Modulation	2.75	No	A Modulation-Domain Loss for Neural-Network-base...	2021-02-15	Code
25	Conv-TasNet-SNR	2.73	No	Exploring the Best Loss Function for DNN-Based L...	2020-08-20	-
26	Sudo rm-rf (U=8)	2.69	No	Continual self-training with bootstrapped remixi...	2021-10-19	Code
27	Proposed (0.35)	2.65	No	Weighted Speech Distortion Losses for Neural-net...	2020-02-12	Code
28	RemixIT (w Sudo U=32)	2.6	Yes	Continual self-training with bootstrapped remixi...	2021-10-19	Code
29	RemixIT (w Sudo U=32)	2.34	Yes	RemixIT: Continual self-training of speech enhan...	2022-02-17	Code
30	Noisy	1.58	No	-	-	-

#1ZipEnhancer (M)
3.81
PESQ-WB
No paper
#2TF-Locoformer (M)SOTA
3.72
PESQ-WB· 2024-08-06
TF-Locoformer: Transformer with Local Modeling by Convolution for Speech Separation and Enhancement Code
#3ZipEnhancer (S)
3.69
PESQ-WB
No paper
#4MambAttention
3.671
PESQ-WB· 2025-07-01
MambAttention: Mamba with Multi-Head Attention for Generalizable Single-Channel Speech Enhancement Code
#5MP-SENetSOTA
3.62
PESQ-WB· 2023-08-17
Explicit Estimation of Magnitude and Phase Spectra in Parallel for High-Quality Speech Enhancement Code
#6xLSTM-SENet
3.588
PESQ-WB· 2025-07-01
MambAttention: Mamba with Multi-Head Attention for Generalizable Single-Channel Speech Enhancement Code
#7BSRNN-S + MRSDSOTA
3.53
PESQ-WB· 2022-12-01
High Fidelity Speech Enhancement with Band-split RNN Code
#8BSRNN-16k
3.45
PESQ-WB· 2022-12-01
High Fidelity Speech Enhancement with Band-split RNN Code
#9MFNET
3.43
PESQ-WB· 2023-06-07
A Mask Free Neural Network for Monaural Speech Enhancement Code
#10BSRNN-S
3.42
PESQ-WB· 2022-12-01
High Fidelity Speech Enhancement with Band-split RNN Code
#11BSRNN
3.32
PESQ-WB· 2022-12-01
High Fidelity Speech Enhancement with Band-split RNN Code
#12CleanUNet-2
3.262
PESQ-WB· 2023-09-12
CleanUNet 2: A Hybrid Speech Denoising Model on Waveform and Spectrogram
#13FRCRNSOTA
3.23
PESQ-WB· 2021-02-03
Monaural Speech Enhancement with Complex Convolutional Block Attention Module and Joint Time Frequency Losses Code
#14FullSubNet+
3.218
PESQ-WB· 2022-03-23
FullSubNet+: Channel Attention FullSubNet with Complex Spectrograms for Speech Enhancement Code
#15CleanUNet
3.146
PESQ-WB· 2022-02-15
Speech Denoising in the Waveform Domain with Self-Attention Code
#16aTENNuate
2.98
PESQ-WB· 2024-09-05
aTENNuate: Optimized Real-time Speech Enhancement with Deep SSMs on Raw Audio
#17Sudo rm -rf (U=32)
2.95
PESQ-WB· 2022-02-17
RemixIT: Continual self-training of speech enhancement models via bootstrapped remixing Code
#18DCTCRN-P
2.82
PESQ-WB
No paper
#19DCTCRN-T
2.82
PESQ-WB
No paper
#20DCCRN-E
2.79
PESQ-WB
No paper
#21PoCoNetSOTA
2.7885
PESQ-WB· 2020-08-11
PoCoNet: Better Speech Enhancement with Frequency-Positional Embeddings, Semi-Supervised Conversational Data, and Biased Loss
#22FullSubNet
2.777
PESQ-WB· 2020-10-29
FullSubNet: A Full-Band and Sub-Band Fusion Model for Real-Time Single-Channel Speech Enhancement Code
#23DCTCRN-S
2.77
PESQ-WB
No paper
#24RNN-Modulation
2.75
PESQ-WB· 2021-02-15
A Modulation-Domain Loss for Neural-Network-based Real-time Speech Enhancement Code
#25Conv-TasNet-SNR
2.73
PESQ-WB· 2020-08-20
Exploring the Best Loss Function for DNN-Based Low-latency Speech Enhancement with Temporal Convolutional Networks
#26Sudo rm-rf (U=8)
2.69
PESQ-WB· 2021-10-19
Continual self-training with bootstrapped remixing for speech enhancement Code
#27Proposed (0.35)SOTA
2.65
PESQ-WB· 2020-02-12
Weighted Speech Distortion Losses for Neural-network-based Real-time Speech Enhancement Code
#28RemixIT (w Sudo U=32)
2.6
PESQ-WB· Extra Data· 2021-10-19
Continual self-training with bootstrapped remixing for speech enhancement Code
#29RemixIT (w Sudo U=32)
2.34
PESQ-WB· Extra Data· 2022-02-17
RemixIT: Continual self-training of speech enhancement models via bootstrapped remixing Code
#30Noisy
1.58
PESQ-WB
No paper