Research map / Vision-language and multimodal / Cross-modal learning
CogCM: Cognition-Inspired Contextual Modeling for Audio-Visual Speech Enhancement
Feixiang Wang, Shuang Yang, Shiguang Shan, Xilin Chen
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 pre-trained noisy audio-visual semantic context module (SeCMPAV) (ablation) | CogCM · LRS3+DNS -15dB | SDR: 5.171 -> 6.680 | Improved |
| add visual-only semantic module trained from scratch (SeCMV) (ablation) | CogCM · LRS3+DNS -15dB | SDR: 5.171 -> 5.414 | Improved |
| use pre-trained audio-visual module on noisy speech instead of clean-pretrained visual module (ablation) | SeCM · LRS3+DNS -15dB | SDR: 6.112 -> 6.680 | Improved |
| use Bi-Mamba signal context module instead of Conformer (ablation) | CogCM (audio-visual) · LRS3+DNS -15dB | SDR: 4.656 -> 5.414 | Improved |
| use SSGM guidance instead of simple addition fusion (ablation) | CogCM with SeCMPAV · LRS3+DNS -15dB | SDR: 6.680 -> 7.284 | Improved |
| use SSGM instead of addition at small PESQ/STOI scale (ablation) | CogCM with SeCMV · LRS3+DNS -5dB | STOI: 0.891 -> 0.892 | No clear effect |
| add frequency upsampling and frequency guidance block (visual frequency) (ablation) | SSGM · LRS3+DNS -15dB | SDR: 6.803 -> 7.284 | Improved |
| model time first then frequency instead of parallel or reverse order (ablation) | SSGM · LRS3+DNS | performance | Improved |
| replace DualAVSE with CogCM (average over SNRs) | AVSE · LRS3+DNS | PESQ: +53.6% over DualAVSE | Improved |
| replace EANet with CogCM | AVSE · GRID/TCD-TIMIT + NoiseX-92 | PESQi: +21.2% PESQi, +35.7% STOIi, +28.7% SI-SDRi | 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.
- VisualVoice — VisualVoice: Audio-Visual Speech Separation With Cross-Modal Consistency (CVPR 2021) · 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
- AVI-Bench: Toward Human-like Audio-Visual Intelligence of Omni-MLLMs (ICML 2026 · similarity 0.87)
- HAVE-Bench: Hierarchical Audio-Visual Evaluation from Perception to Interaction (CVPR 2026 · similarity 0.85)
- From Senses to Decisions: The Information Flow of Auditory and Visual Perception in Multimodal LLMs (NeurIPS 2026 · similarity 0.85)
- SVHalluc: Benchmarking Speech-Vision Hallucination in Audio-Visual Large Language Models (CVPR 2026 · similarity 0.85)
- video-SALMONN: Speech-Enhanced Audio-Visual Large Language Models (ICML 2024 · similarity 0.85)
- Visual Hallucination Elevates Speech Recognition (AAAI 2024 · similarity 0.84)
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.
