Speech Enhancement on VoiceBank + DEMAND

42개 결과 · ⬇ CSV · JSON

PESQ (wb)

Rank	Model	PESQ (wb)	CBAK	COVL	CSIG	STOI	ESTOI	SSNR	SI-SDR	Paper	Code	Year
1	ROSE-CD(PESQ)	3.99	3.37	4.30	4.63	92.6	0.83	0.927	0.40	Robust One-step Speech Enhancement via Consistency Distillation	LiangXu123/Robust-One-step-Speech-Enhancement-via-Consistency-Distillation-ROSE-CD-	2025
2	PESQetarian	3.82	2.49	3.5	3.63	0.92	0.84	-2.72	-19.8	The PESQetarian: On the Relevance of Goodhart's Law for Speech Enhancement		2024
3	Mamba-SEUNet L (+PCS)	3.73	3.67	4.40	4.82	96	–	–	–	Mamba-SEUNet: Mamba UNet for Monaural Speech Enhancement	MyParadise21/Mamba-SEUNet	2024
4	Schrödinger bridge (PESQ loss)	3.70	–	–	–	–	–	–	–	Investigating Training Objectives for Generative Speech Enhancement	sp-uhh/sgmse	2024
5	SEMamba (+PCS)	3.69	3.63	4.37	4.79	96	–	–	–	An Investigation of Incorporating Mamba for Speech Enhancement	roychao19477/semamba	2024
6	ZipEnhancer (S, \lamba_6 = 0)	3.63	3.87	4.36	4.81	96.19	–	8.33	19.09
7	PrimeK-Net	3.61	3.98	4.35	4.81	96	–	–	–	PrimeK-Net: Multi-scale Spectral Learning via Group Prime-Kernel Convolutional Neural Networks for Single Channel Speech Enhancement	huaidanquede/PrimeK-Net	2025
7	ZipEnhancer (S, \lamba_6 = 0.2)	3.61	3.97	4.35	4.81	96.22	–	10.01	19.96
9	MP-SENet	3.60	3.99	4.34	4.81	0.96	–	–	–	Explicit Estimation of Magnitude and Phase Spectra in Parallel for High-Quality Speech Enhancement	yxlu-0102/MP-SENet	2023
10	PCS_CS_WAVLM	3.54	3.49	4.20	4.75	0.96	–	–	–
11	xLSTM-SENet2	3.53	3.98	4.27	4.78	0.96	–	–	–	xLSTM-SENet: xLSTM for Single-Channel Speech Enhancement	nikolaikyhne/xlstm-senet	2025
12	SCP-CMGAN	3.52	3.97	4.25	4.75	96	–	10.82	–	SCP-GAN: Self-Correcting Discriminator Optimization for Training Consistency Preserving Metric GAN on Speech Enhancement Tasks		2022
13	ROSE-CD	3.49	3.33	4.04	4.523	94.73	0.87	3.34	17.80	Robust One-step Speech Enhancement via Consistency Distillation	LiangXu123/Robust-One-step-Speech-Enhancement-via-Consistency-Distillation-ROSE-CD-	2025
14	D2Former	3.43	–	–	–	–	–	–	–	Monaural Speech Enhancement with Complex Convolutional Block Attention Module and Joint Time Frequency Losses	modelscope/ClearerVoice-Studio · alibabasglab/frcrn	2021
15	CMGAN	3.41	3.94	4.12	4.63	96	–	11.1	–	CMGAN: Conformer-Based Metric-GAN for Monaural Speech Enhancement	ruizhecao96/cmgan · SherifAbdulatif/CMGAN	2022
16	PCS	3.35	–	3.92	4.43	95	–	–	–	Perceptual Contrast Stretching on Target Feature for Speech Enhancement	roychao19477/pcs	2022
17	D²Net	3.27	3.18	3.92	4.63	96	–	–	–	D²Net: A Denoising and Dereverberation Network Based on Two-branch Encoder and Dual-path Transformer		2022
17	aTENNuate	3.27	2.85	3.96	4.57	–	–	–	15.04	aTENNuate: Optimized Real-time Speech Enhancement with Deep SSMs on Raw Audio		2024
19	Centaurus (0.51M)	3.25	–	–	–	–	–	–	–	Let SSMs be ConvNets: State-space Modeling with Optimal Tensor Contractions		2025
20	MetricGAN-OKD	3.24	3.07	3.73	4.23	–	–	–	–	MetricGAN-OKD: Multi-Metric Optimization of MetricGAN via Online Knowledge Distillation for Speech Enhancement	wooseok-shin/MetricGAN-OKD	2023