mlsanityData to 19 Oct 2026

ArrayDPS-Refine: Generative Refinement of Discriminative Multi-Channel Speech Enhancement

arXiv 2603.24385 ↗First posted 25 Mar 2026eess.AScited by 0

On the map: Speech and audio language models → Speech enhancement and noise suppression

Abstract (arXiv)

Multi-channel speech enhancement aims to recover clean speech from noisy multi-channel recordings. Most deep learning methods employ discriminative training, which can lead to non-linear distortions from regression-based objectives, especially under challenging environmental noise conditions. Inspired by ArrayDPS for unsupervised multi-channel source separation, we introduce ArrayDPS-Refine, a method designed to enhance the outputs of discriminative models using a clean speech diffusion prior. ArrayDPS-Refine is training-free, generative, and array-agnostic. It first estimates the noise spatial covariance matrix (SCM) from the enhanced speech produced by a discriminative model, then uses this estimated noise SCM for diffusion posterior sampling. This approach allows direct refinement of any discriminative model's output without retraining. Our results show that ArrayDPS-Refine consistently improves the performance of various discriminative models, including state-of-the-art waveform and STFT domain models. Audio demos are provided at https://xzwy.github.io/ArrayDPSRefineDemo/.

Key facts. No citing papers in the map yet (data through 19 Oct 2026). Placed in Speech enhancement and noise suppression, one of 11 directions in Speech and audio language models. Builds on 28 of its 39 references that are in the map. First posted on arXiv 25 Mar 2026.

Builds on

28 of 39 references.

Closest ancestors · 10

Same direction, newest first

  1. 2024Rethinking Processing Distortions: Disentangling the Impact of Speech Enhancement Errors on Speech Recognition Performance23 Apr
  2. 2024Improving Design of Input Condition Invariant Speech Enhancement25 Jan
  3. 2023uSee: Unified Speech Enhancement and Editing with Conditional Diffusion Models2 Oct
  4. 2023Toward Universal Speech Enhancement for Diverse Input Conditions29 Sep
  5. 2022StoRM: A Diffusion-based Stochastic Regeneration Model for Speech Enhancement and Dereverberation22 Dec
  6. 2022LA-VocE: Low-SNR Audio-visual Speech Enhancement using Neural Vocoders20 Nov
  7. 2022Diffiner: A Versatile Diffusion-based Generative Refiner for Speech Enhancement27 Oct
  8. 2022Speech Enhancement and Dereverberation with Diffusion-based Generative Models11 Aug
  9. 2022Conditional Diffusion Probabilistic Model for Speech Enhancement10 Feb
  10. 2021TPARN: Triple-path Attentive Recurrent Network for Time-domain Multichannel Speech Enhancement20 Oct

Other references · 18

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,267
  2. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,295
  3. 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,661
  4. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,783
  5. 2022Diffusion Posterior Sampling for General Noisy Inverse Problemscited by 1,431
  6. 2022Denoising Diffusion Restoration Modelscited by 965
  7. 2018SDR - half-baked or well done?cited by 626
  8. 2020DNSMOS: A Non-Intrusive Perceptual Objective Speech Quality metric to evaluate Noise Suppressorscited by 356
  9. 2020The INTERSPEECH 2020 Deep Noise Suppression Challenge: Datasets, Subjective Testing Framework, and Challenge Resultscited by 230
  10. 2017Pyroomacoustics: A Python package for audio room simulations and array processing algorithmscited by 175
  11. 2024The T05 System for The VoiceMOS Challenge 2024: Transfer Learning from Deep Image Classifier to Naturalness MOS Prediction of High-Quality Synthetic Speechcited by 118
  12. 2019End-to-end Microphone Permutation and Number Invariant Multi-channel Speech Separationcited by 70
6 more
  1. 2021Convolutive Prediction for Monaural Speech Dereverberation and Noisy-Reverberant Speaker Separationcited by 21
  2. 2023EDMSound: Spectrogram Based Diffusion Models for Efficient and High-Quality Audio Synthesiscited by 16
  3. 2025ArrayDPS: Unsupervised Blind Speech Separation with a Diffusion Priorcited by 10
  4. 2021Time-domain Ad-hoc Array Speech Enhancement Using a Triple-path Networkcited by 7
  5. 2023Self-supervised learning with diffusion-based multichannel speech enhancement for speaker verification under noisy conditionscited by 2
  6. 2025Unsupervised Multi-channel Speech Dereverberation via Diffusioncited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Unified Diffusion Refinement for Multi-Channel Speech Enhancement and Separation
  2. 2025MDDM: A Multi-view Discriminative Enhanced Diffusion-based Model for Speech Enhancement
  3. 2024Diffusion-based Generative Modeling with Discriminative Guidance for Streamable Speech Enhancement
  4. 2022Analysing Diffusion-based Generative Approaches versus Discriminative Approaches for Speech Restoration
  5. 2023DDTSE: Discriminative Diffusion Model for Target Speech Extraction
  6. 2025A Composite Predictive-Generative Approach to Monaural Universal Speech Enhancement
  7. 2026VM-ARRAYDPS: Virtual Microphone Augmented Diffusion Posterior Sampling for Unsupervised Blind Speech Separation
  8. 2024Noise-robust Speech Separation with Fast Generative Correction