What question did this study set out to answer?

This research aims to explore the use of virtual agents for training gesture recognition algorithms while addressing data limitations.

January 18, 2026Behavior Research MethodsOpen Access

Virtual agents as a scalable tool for diverse, robust gesture recognition

View Full Paper

Ask AI

Bookmark

Authors

Virtual agents train recognition algorithms to improve gesture detection accuracy, suggesting new research efficiencies.

This research aims to explore the use of virtual agents for training gesture recognition algorithms while addressing data limitations.
Developed a virtual agent using motion capture (mocap) data to create various training scenarios.
Customized lighting, backgrounds, and camera angles for diverse datasets.
Trained a gesture recognition algorithm on the virtual agent dataset and tested its accuracy on human images.
Achieved 85.9% accuracy in optimal conditions with the virtual agent.
Accuracy dropped to 71.6% with background clutter and reduced lighting.
Target handshape classification accuracy ranged from 72% to 95% when tested on human images.

Loy et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark