A real-time gesture recognition system using near-infrared imagery.

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Machine Learning

Background:

• Hand gesture recognition is key for intuitive human-computer interaction (HCI).
• Current skeleton-based methods limit gesture recognition due to difficulties in robust skeleton extraction for all poses.
• Near-infrared (NIR) imaging offers potential for gesture analysis.

Purpose of the Study:

• To develop a real-time hand gesture recognition system that directly analyzes NIR imagery.
• To overcome the limitations of skeleton-based approaches by avoiding skeleton information extraction.
• To expand the range of recognizable static and dynamic hand gestures.

Main Methods:

• A novel system for real-time hand gesture recognition using a near-infrared device.
• Direct analysis of infrared imagery, bypassing the need for skeleton data extraction.
• Development and utilization of a new dataset comprising near-infrared hand gesture imagery.

Main Results:

• The proposed system successfully infers static and dynamic gestures directly from NIR imagery.
• Achieved superior performance compared to existing state-of-the-art methods on the newly created dataset.
• Demonstrated the capability to recognize a significantly wider range of hand gestures than skeleton-based systems.

Conclusions:

• Direct analysis of NIR imagery offers a more robust and comprehensive approach to hand gesture recognition.
• The developed system expands the possibilities for immersive and intuitive human-computer interaction.
• The findings suggest a promising new direction for gesture recognition technology, particularly in challenging environments.