Related Experiment Video
Updated: Aug 6, 2026
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
A Meta-Learning Framework for Few-Shot Dynamic Hand Gesture Recognition with Soft Temporal-aware Contrastive Learning
IEEE Journal of Biomedical and Health Informatics
|July 23, 2026
Summary
Few-shot learning for surface electromyography (sEMG) gesture recognition is enhanced by STC+Meta. This framework improves data efficiency and personalization, achieving high accuracy with minimal subject-specific data.
Area of Science:
- Biomedical Engineering
- Machine Learning
- Human-Computer Interaction
Background:
- Few-shot learning enhances data efficiency in surface electromyography (sEMG) gesture recognition.
- Cross-subject and cross-posture generalization in sEMG recognition is hindered by nonstationary signals, electrode displacement, and time-varying dynamics.
- Existing methods struggle with efficient personalization for new subjects or conditions.
Purpose of the Study:
- To propose STC+Meta, a framework coupling temporally informed representation learning with efficient personalization for sEMG gesture recognition.
- To improve data efficiency and generalization across subjects and postures in sEMG-based gesture recognition.
- To enable calibration-efficient personalization using limited subject-specific data.
Main Methods:
- Developed a framework (STC+Meta) combining soft temporal contrastive (STC) pre-training and implicit MAML-based meta-adaptation.
- STC pre-training uses a soft temporal contrastive loss to encourage temporal continuity and preserve discriminative transients from sEMG and accelerometer (ACC) data.
- Implicit MAML provides data-efficient meta-updates for personalization using only 10% of subject-specific calibration data.
Main Results:
- Achieved 97.63% accuracy in a 7-class gesture recognition task using 10% subject-specific data (∼28 s) for calibration.
- Attained 95.75% accuracy in a 10-class task-transfer setting.
- STC+Meta demonstrated faster convergence and lower inter-subject variability compared to batch fine-tuning with limited calibration data.
Conclusions:
- Temporally aware contrastive pre-training combined with meta-learning enables highly efficient personalization for myoelectric gesture recognition.
- The STC+Meta framework significantly reduces the calibration data requirement for accurate sEMG gesture recognition.
- This approach addresses key challenges in real-world sEMG applications, particularly under biased conditions.
Related Concept Videos
01:12Observational Learning
Albert Bandura's observational learning, also known as imitation or modeling, occurs when a person observes and imitates another's behavior. It is a quicker process than operant conditioning. A well-known example is the Bobo doll study, where children who saw an adult acting aggressively towards the doll were more likely to act aggressively when left alone, compared to those who observed a nonaggressive adult. Many psychologists view observational learning as a form of latent learning because...
01:18Introduction to Learning
Learning is the process of acquiring knowledge or skills through practice or experience, leading to long-lasting behavioral changes. This acquisition occurs through interaction with the environment and requires practice or experience. For instance, mastering a skill such as surfing requires considerable practice and experience, highlighting the essential role of repeated interactions with the environment in learning.
In contrast to learned behaviors, unlearned behaviors such as crying, sexual...
In contrast to learned behaviors, unlearned behaviors such as crying, sexual...