MyoSem: Aligning Electromyography with Natural-Language Action Semantics
MyoSem is an alignment framework that connects raw muscle activity signals (EMG) with natural-language descriptions of hand actions. It achieves this by creating a shared semantic space using multi-view text descriptions and an activation-aware encoder for EMG. The system allows bidirectional retrieval, meaning you can search for a specific action description to find matching muscle signals.