Research map / Image generation / Fast diffusion sampling
DOSE: Diffusion Dropout with Adaptive Prior for Speech Enhancement
Wenxin Tai, Yue Lei, Fan Zhou, Goce Trajcevski, Ting Zhong
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 |
|---|---|---|---|
| Use handpicked instead of fixed timesteps for two-step generation (ablation) | DOSE · VB matched | PESQ: 2.48 -> 2.56 | Improved |
| Use handpicked instead of fixed timesteps for two-step generation (ablation) | DOSE · CHIME-4 mismatched | PESQ: 1.44 -> 1.52 | Improved |
| Use handpicked instead of fixed timesteps for two-step generation (ablation) | DOSE · CHIME-4 mismatched | STOI(%): 83.6 -> 86.6 | Improved |
| Use two steps instead of one step (handpicked timesteps) (ablation) | DOSE · VB matched | PESQ: 2.50 -> 2.56 | Improved |
| Use two steps instead of one step (handpicked timesteps) (ablation) | DOSE · CHIME-4 mismatched | STOI(%): 86.4 -> 86.6 | No clear effect |
| Use two steps instead of one step (fixed timesteps) (ablation) | DOSE · CHIME-4 mismatched | PESQ: 1.38 -> 1.44 | Improved |
| Use DOSE over CDiffuSE with fewer steps | DOSE vs CDiffuSE · CHIME-4 mismatched | STOI(%): 82.8 -> 86.6 | Improved |
| Use DOSE over CDiffuSE | DOSE vs CDiffuSE · CHIME-4 mismatched | PESQ: 1.58 -> 1.52 | Hurt |
| Use DOSE over DiffWave (deterministic) | DOSE vs DiffWave · CHIME-4 mismatched | STOI(%): 72.3 -> 86.6 | Improved |
| Use DOSE over DiffWave (deterministic) | DOSE vs DiffWave · VB matched | PESQ: 2.51 -> 2.56 | Improved |
| Use DOSE over DR-DiffuSE in subjective naturalness, mismatched | DOSE vs DR-DiffuSE · Subjective MOS mismatched | MOS: 2.48 -> 3.48 | Improved |
| Use mild condition factor instead of estimated speech as adaptive prior (ablation) | unconditional diffusion · CHIME-4 mismatched | PESQ | Improved |
Later papers that compared with this method
1 later accepted paper reports DOSE+ as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Dual-View Predictive Diffusion: Lightweight Speech Enhancement via Spectrogram-Image Synergy (ICML 2026 · better on 1 of 1 shared result)
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.
- DiffWave — DiffWave: A Versatile Diffusion Model for Audio Synthesis (ICLR 2021) · better on 4 of 4 shared results
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
- Shallow Diffusion for Fast Speech Enhancement (Student Abstract) (AAAI 2024 · similarity 0.89)
- Predictive–Generative Drift Decomposition for Speech Enhancement and Separation (NeurIPS 2026 · similarity 0.87)
- RestoreGrad: Signal Restoration Using Conditional Denoising Diffusion Models with Jointly Learned Prior (ICML 2025 · similarity 0.86)
- Is Noise Conditioning Necessary for Denoising Generative Models? (ICML 2025 · similarity 0.86)
- Complex-Cycle-Consistent Diffusion Model for Monaural Speech Enhancement (AAAI 2025 · similarity 0.85)
- DDDM-VC: Decoupled Denoising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust Voice Conversion (AAAI 2024 · 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.
