MyoFlow improves muscle signal gesture recognition across users and sessions
MyoFlow: Anchor-Tied Rectified Flow for HD-sEMG Gesture Recognition Across Sessions and Subjects
Machine Learning
Summary
Recognizing hand gestures from muscle signals can be tricky because the signals change when people wear sensors differently or over time. The authors developed MyoFlow, a method that more accurately identifies gestures from these muscle signals even when conditions vary. It works by moving signal representations closer to known gesture patterns, allowing it to predict gestures without extra steps. Tests showed it outperforms previous methods, achieving high accuracy across different days and users.
What this means in practice
- •For prosthetic device developers: Improve the accuracy and reliability of hand gesture recognition in prosthetic control systems across multiple users and sessions without retraining classifiers.$Commercial implications: Enables more user-friendly muscle-controlled prosthetics with consistent gesture recognition across time and users, suitable for sale to medical device manufacturers.
- •For assistive robotics engineers: Enhance robotic assistance devices by better interpreting muscle signals for natural gesture-based control that adapts across different users and times.
Authors
Chenhao Wu, Dingjie Peng, Satoshi Funabashi, Satoshi Konishi, Wuqiang Yang, Hiroshi Onoda, Hironori Washizaki, Jiang Liu
Abstract
High-density surface electromyography (HD-sEMG) gesture recognition supports prosthetic control, assistive robotics, and rehabilitation, but electrode re-donning and physiological variability cause distribution shifts that degrade accuracy across sessions and subjects. Generative HD-sEMG models primarily synthesize signals for augmentation; although diffusion models enhance representation learning, prediction still relies on a separate classifier. To tie learned dynamics to the decision rule, we propose MyoFlow, the first discriminative flow-matching framework for HD-sEMG recognition across sessions and subjects. It recasts classification as anchor-tied transport: a domain-conditioned rectified flow moves encoded windows toward gesture anchors that serve as transport targets and define the nearest-anchor decision geometry, enabling zero-shot prediction without an independent head. On the Hyser dataset, MyoFlow improves mean cross-session and cross-subject accuracy over the strongest diffusion-based baseline by 4.24\% and 6.37\%, respectively, and achieves 91.71\% mean zero-shot accuracy and 97.39\% mean few-shot accuracy across multiple days on the CEMHSEY dataset.