Research map / Vision-language and multimodal / Cross-modal learning
EMBridge: Enhancing Gesture Generalization from EMG Signals Through Cross-modal Representation Learning
Wenhui Cui, Christopher Michael Sandino, Hadi Pouransari, Ran Liu, Juri Minxha, Ellen L. Zippi, Erdrin Azemi, Behrooz Mahasseni
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 Q-Former (ablation) | EMBridge · emg2pose unseen, zero-shot | balanced accuracy: 0.494 -> 0.528 | Improved |
| add Q-Former (ablation) | EMBridge · emg2pose in-dist, linear probe | balanced accuracy: 0.793 -> 0.785 | Hurt |
| add masked pose reconstruction loss (ablation) | EMBridge · emg2pose unseen, zero-shot | balanced accuracy: 0.516 -> 0.528 | Improved |
| add CASCLe soft contrastive loss (ablation) | EMBridge · emg2pose unseen, zero-shot | balanced accuracy: 0.509 -> 0.528 | Improved |
| add CASCLe soft contrastive loss (ablation) | EMBridge · emg2pose unseen, linear probe | balanced accuracy: 0.485 -> 0.505 | Improved |
| replace label smoothing with CASCLe (ablation) | EMBridge · emg2pose unseen, zero-shot | balanced accuracy: 0.511 -> 0.528 | Improved |
| replace SoftCLIP with CASCLe (ablation) | EMBridge · emg2pose unseen, zero-shot | balanced accuracy: 0.510 -> 0.528 | Improved |
| add label smoothing 0.1 to InfoNCE | Q-Former · emg2pose unseen, zero-shot | balanced accuracy: 0.498 -> 0.498 | No clear effect |
| increase batch size from 256 to 1024 (ablation) | CPEP · emg2pose unseen, zero-shot | balanced accuracy: 0.481 -> 0.433 | Hurt |
| increase batch size from 256 to 1024 (ablation) | EMBridge · emg2pose unseen, zero-shot | balanced accuracy: 0.528 -> 0.523 | No clear effect |
| increase top-k clusters in CASCLe to 10 or 20 (ablation) | EMBridge · emg2pose | balanced accuracy | Hurt |
| use 16 queries instead of 8 or 32 (ablation) | EMBridge · emg2pose | balanced accuracy | Improved |
Closest papers by meaning
- emg2pose: A Large and Diverse Benchmark for Surface Electromyographic Hand Pose Estimation (NeurIPS 2024 · similarity 0.89)
- emg2qwerty: A Large Dataset with Baselines for Touch Typing using Surface Electromyography (NeurIPS 2024 · similarity 0.87)
- EMGBench: Benchmarking Out-of-Distribution Generalization and Adaptation for Electromyography (NeurIPS 2024 · similarity 0.86)
- SpGesture: Source-Free Domain-adaptive sEMG-based Gesture Recognition with Jaccard Attentive Spiking Neural Network (NeurIPS 2024 · similarity 0.86)
- Revisiting Noise Resilience Strategies in Gesture Recognition: Short-Term Enhancement in sEMG Analysis (ICML 2025 · similarity 0.85)
- SplashNet: Split‑and‑Share Encoders for Accurate and Efficient Typing with Surface Electromyography (NeurIPS 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.
