Pith. sign in

REVIEW 12 cited by

Universal Speech Enhancement with Score-based Diffusion

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2206.03065 v2 pith:OKFGU4AZ submitted 2022-06-07 cs.SD cs.LGeess.AS

Joan Serrà , Santiago Pascual , Jordi Pons , R. Oguz Araz , Davide Scaini This is my paper · ORCID
classification cs.SDcs.LGeess.AS
keywords enhancementspeechapproachdiffusionuniversalbackgroundgenerativenoise
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Removing background noise from speech audio has been the subject of considerable effort, especially in recent years due to the rise of virtual communication and amateur recordings. Yet background noise is not the only unpleasant disturbance that can prevent intelligibility: reverb, clipping, codec artifacts, problematic equalization, limited bandwidth, or inconsistent loudness are equally disturbing and ubiquitous. In this work, we propose to consider the task of speech enhancement as a holistic endeavor, and present a universal speech enhancement system that tackles 55 different distortions at the same time. Our approach consists of a generative model that employs score-based diffusion, together with a multi-resolution conditioning network that performs enhancement with mixture density networks. We show that this approach significantly outperforms the state of the art in a subjective test performed by expert listeners. We also show that it achieves competitive objective scores with just 4-8 diffusion steps, despite not considering any particular strategy for fast sampling. We hope that both our methodology and technical contributions encourage researchers and practitioners to adopt a universal approach to speech enhancement, possibly framing it as a generative task.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 12 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. UniPASE: A Generative Model for Universal Speech Enhancement with High Fidelity and Low Hallucinations

    eess.AS 2026-04 unverdicted novelty 6.0 of 10

    UniPASE extends the PASE framework with DeWavLM-Omni to convert degraded speech into high-fidelity, low-hallucination audio across sampling rates via phonetic enhancement, acoustic adaptation, and multi-rate vocoding.

  2. UniSE: A Unified Framework for Decoder-Only Autoregressive LM-Based Speech Enhancement

    cs.SD 2025-10 conditional novelty 6.0 of 10

    A 63M-parameter decoder-only LM, UniSE, unifies speech restoration, target speaker extraction, and speech separation by generating BiCodec discrete tokens under task-specific prompts.

  3. TalkLess: Blending Extractive and Abstractive Speech Summarization for Editing Speech to Preserve Content and Style

    cs.HC 2025-07 conditional novelty 6.0 of 10

    TalkLess blends extractive and abstractive speech summarization through LLM candidate generation and a weighted scoring function, then converts transcript edits to audio with VoiceCraft, evaluating favorably against a...

  4. TGIF: Talker Group-Informed Familiarization of Target Speaker Extraction

    eess.AS 2025-07 conditional novelty 6.0 of 10

    A knowledge distillation framework adapts a compact target speaker extraction model to a specific talker group, improving over generic models using only unlabeled mixtures.

  5. Posterior Transition Modeling for Unsupervised Diffusion-Based Speech Enhancement

    cs.SD 2025-07 conditional novelty 6.0 of 10

    Explicitly modeling the posterior transition in diffusion-based speech enhancement, including a noisy-speech diffusion that makes the likelihood tractable, improves unsupervised enhancement and robustness to domain shift.

  6. User-guided Generative Source Separation

    cs.SD 2025-07 conditional novelty 6.0 of 10

    GuideSep separates arbitrary target instruments from a mixture using user-provided waveform mimicry and mel-spectrogram masks, and outperforms a same-architecture mask-prediction baseline in SDR and listening tests.

  7. A Composite Predictive-Generative Approach to Monaural Universal Speech Enhancement

    eess.AS 2025-05 conditional novelty 6.0 of 10

    PGUSE combines a predictive speech enhancer with a diffusion model, fusing their outputs and truncating the diffusion start to improve universal speech enhancement with low inference cost.

  8. Interspeech 2025 URGENT Speech Enhancement Challenge

    eess.AS 2025-05 conditional novelty 6.0 of 10

    A 32-team speech enhancement challenge found that the top-scoring system is discriminative, but generative systems win subjective listening and can severely mis-transcribe Chinese and Japanese audio.

  9. SEMamba++: A General Speech Restoration Framework Leveraging Global, Local, and Periodic Spectral Patterns

    eess.AS 2026-03 conditional novelty 5.5 of 10

    SEMamba++ combines Frequency GLP (FAN-based global-periodic + local conv) with multi-resolution parallel TFDP and learnable softplus mapping to outperform GSR baselines on VCTK, URGENT and AATC while remaining efficient.

  10. Voice-ENHANCE: Speech Restoration using a Diffusion-based Voice Conversion Framework

    cs.SD 2025-05 conditional novelty 5.0 of 10

    A diffusion voice conversion model, conditioned on clean speaker embeddings and HuBERT content features, is applied after a generative speech restorer to achieve state-of-the-art-comparable speech quality.

  11. Few-step Adversarial Schr\"{o}dinger Bridge for Generative Speech Enhancement

    cs.SD 2025-06 conditional novelty 4.0 of 10

    Adding an adversarial GAN objective to a Schrödinger Bridge speech enhancement model enables high-quality denoising and dereverberation at one to four sampling steps, surpassing slower baselines on full-band benchmarks.

  12. TS-URGENet: A Three-stage Universal Robust and Generalizable Speech Enhancement Network

    eess.AS 2025-05 conditional novelty 4.0 of 10

    TS-URGENet, a filling-separation-restoration cascade with metric-aware fine-tuning, places 2nd in the URGENT 2025 universal speech enhancement challenge.

Pith tools