Research map / Vision-language and multimodal / Human and embodied modeling
Listening to the World Improves Speech Command Recognition
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 |
|---|---|---|---|
| Pretrain DenseNet121 on UrbanSound8K before speech-command training | DenseNet121 without multiscale input · Speech Commands left vs. right | accuracy: 89.19 -> 91.40 | Helped |
| Pretrain DenseNet121 on UrbanSound8K before speech-command training | DenseNet121 without multiscale input · Speech Commands 20 Command Terms | accuracy: 81.32 -> 82.48 | Helped |
| Pretrain DenseNet121 on UrbanSound8K before speech-command training | DenseNet121 without multiscale input · Speech Commands All 30 Terms | accuracy: 80.13 -> 81.55 | Helped |
| Add multiscale dilated-convolution input (ablation) | DenseNet121 with fresh initialization · Speech Commands left vs. right | accuracy: 89.19 -> 88.54 | Hurt |
| Add multiscale dilated-convolution input (ablation) | DenseNet121 with fresh initialization · Speech Commands 20 Command Terms | accuracy: 81.32 -> 82.22 | Helped |
| Add multiscale dilated-convolution input (ablation) | DenseNet121 with fresh initialization · Speech Commands All 30 Terms | accuracy: 80.13 -> 82.11 | Helped |
| Pretrain on UrbanSound8K and add multiscale dilated-convolution input | DenseNet121 · Speech Commands left vs. right | accuracy: 89.19 -> 95.32 | Helped |
| Pretrain on UrbanSound8K and add multiscale dilated-convolution input | DenseNet121 · Speech Commands 20 Command Terms | accuracy: 81.32 -> 85.52 | Helped |
| Pretrain on UrbanSound8K and add multiscale dilated-convolution input | DenseNet121 · Speech Commands All 30 Terms | accuracy: 80.13 -> 84.35 | Helped |
| Add multiscale dilated-convolution input (ablation) | DenseNet121 · UrbanSound8K | accuracy: 70.61 -> 71.04; 72.53 -> 71.27; 70.87 -> 71.86 | Mixed |
| Use DenseNet rather than ResNet or SB-CNN | Audio spectrogram classification · UrbanSound8K | accuracy: 70.61 vs. 69.45 | Helped |
| Use pretrained multiscale models instead of fresh models | DenseNet121 speech-command training · Speech Commands left vs. right | training data needed for similar accuracy: 40% | Helped |
Closest papers by meaning
- Breaking Adversarial Transferability in Fine-Tuned Speech Recognition (NeurIPS 2026 · similarity 0.84)
- AND: Audio Network Dissection for Interpreting Deep Acoustic Models (ICML 2024 · similarity 0.84)
- Measuring Audio's Impact on Correctness: Audio-Contribution-Aware Post-Training of Large Audio Language Models (ICLR 2026 · similarity 0.84)
- Massive Sound Embedding Benchmark (MSEB) (NeurIPS 2025 · similarity 0.83)
- UniAudio 1.5: Large Language Model-Driven Audio Codec is A Few-Shot Audio Task Learner (NeurIPS 2024 · similarity 0.83)
- Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputs (NeurIPS 2024 · 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.