Research map / Image generation / Image restoration
Dual-View Predictive Diffusion: Lightweight Speech Enhancement via Spectrogram-Image Synergy
Ke Xue, Rongfei Fan, Kai Li, Shanping Yu, Puning Zhao, Jianping An
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Training-free Lossless Boost (TLB) at inference (ablation) | DVPD · WSJ0-UNI | PESQ: 2.99 -> 3.15 | Improved |
| remove LISA module (ablation) | DVPD · WSJ0-UNI | PESQ: 2.99 -> 2.71 | Improved |
| add FANC encoder (ablation) | DVPD · WSJ0-UNI | PESQ: 2.93 -> 2.99 | Improved |
| add Frequency-aware Interaction module (ablation) | DVPD · WSJ0-UNI | PESQ: 2.91 -> 2.99 | Improved |
| add anti-wrapping phase loss (ablation) | DVPD · WSJ0-UNI | PESQ: 2.91 -> 2.99 | Improved |
| use BBED SDE instead of OUVE SDE (ablation) | DVPD · WSJ0-UNI | PESQ: 2.89 -> 2.99 | Improved |
| use predicted phase from predictive branch instead of degraded phase (ablation) | DVPD · WSJ0-UNI | PESQ: 2.65 -> 2.99 | Improved |
| apply TLB to low-quality samples (PESQ<2) (ablation) | DVPD · WSJ0-UNI | PESQ: 1.73 -> 1.95 | Improved |
| apply TLB to high-quality samples (PESQ>=3) (ablation) | DVPD · WSJ0-UNI | PESQ: 3.20 -> 3.24 | Improved |
| apply TLB (SI-SDR effect) (ablation) | DVPD · WSJ0-UNI (PESQ<2) | SI-SDR: 16.99 -> 15.85 | Hurt |
| apply TLB to other diffusion baseline PGUSE (U-Net) | PGUSE (U-Net) · WSJ0-UNI | PESQ: 2.82 -> 3.02 | Improved |
| apply TLB to other diffusion baseline StoRM | StoRM · WSJ0-UNI | PESQ: 2.75 -> 2.81 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DOSE+ — DOSE: Diffusion Dropout with Adaptive Prior for Speech Enhancement (NeurIPS 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Predictive–Generative Drift Decomposition for Speech Enhancement and Separation (NeurIPS 2026 · similarity 0.88)
- Images that Sound: Composing Images and Sounds on a Single Canvas (NeurIPS 2024 · similarity 0.87)
- MDSGen: Fast and Efficient Masked Diffusion Temporal-Aware Transformers for Open-Domain Sound Generation (ICLR 2025 · similarity 0.86)
- Shallow Diffusion for Fast Speech Enhancement (Student Abstract) (AAAI 2024 · similarity 0.86)
- Rethinking Training Targets, Architectures and Data Quality for Universal Speech Enhancement (NeurIPS 2026 · similarity 0.86)
- Complex-Cycle-Consistent Diffusion Model for Monaural Speech Enhancement (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
