Neuromotor Hierarchy Network: Physiological Inductive Biases for Robust Generalization in sEMG Decoding
The Neuromotor Hierarchy Network (NHN) uses a physiology-guided sEMG decoding architecture to infer a compact, task-aligned latent neuromotor state. It learns this state from task supervision to improve generalization across different users and sessions. This approach outperforms existing methods by leveraging inherent neuromuscular organization.