Research map / Label-efficient and robust learning / Recognition and long-tail
PHASEN: A Phase-and-Harmonics-Aware Speech Enhancement Network
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the two-stream architecture instead of a single stream with cIRM supervision (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): +0.85dB | Helped |
| Use the two-stream architecture instead of a single stream with cIRM supervision (ablation) | PHASEN · AVSpeech + AudioSet | PESQ: +0.42 | Helped |
| Add frequency transformation blocks instead of 5×5 convolutions (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): +0.74 dB | Helped |
| Add frequency transformation blocks instead of 5×5 convolutions (ablation) | PHASEN · AVSpeech + AudioSet | PESQ: +0.09 | Helped |
| Add amplitude-to-phase communication (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): +0.49dB | Helped |
| Add amplitude-to-phase communication (ablation) | PHASEN · AVSpeech + AudioSet | PESQ: +0.05 | Helped |
| Add phase-to-amplitude communication (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): +0.22dB | Helped |
| Add phase-to-amplitude communication (ablation) | PHASEN · AVSpeech + AudioSet | PESQ: +0.02 | Helped |
| Add ReLU or Tanh after intermediate phase-stream convolutional layers (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): 0.02dB-0.16dB drop | Hurt |
| Add ReLU after the last phase-stream convolutional layer (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): 5.52dB drop | Hurt |
| Use gLN instead of BN in the amplitude stream (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): 0.97dB drop | Hurt |
| Use BN instead of gLN in the phase stream (ablation) | PHASEN · AVSpeech + AudioSet | SDR (dB): 0.09dB drop | Hurt |
Closest papers by meaning
- BSDB-Net: Band-Split Dual-Branch Network with Selective State Spaces Mechanism for Monaural Speech Enhancement (AAAI 2025 · similarity 0.85)
- Complex-Cycle-Consistent Diffusion Model for Monaural Speech Enhancement (AAAI 2025 · similarity 0.84)
- FINALLY: fast and universal speech enhancement with studio-like quality (NeurIPS 2024 · similarity 0.84)
- DTF-AT: Decoupled Time-Frequency Audio Transformer for Event Classification (AAAI 2024 · similarity 0.84)
- Rethinking Training Targets, Architectures and Data Quality for Universal Speech Enhancement (NeurIPS 2026 · similarity 0.83)
- MDSGen: Fast and Efficient Masked Diffusion Temporal-Aware Transformers for Open-Domain Sound Generation (ICLR 2025 · similarity 0.83)
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.