Simulation-driven design of smart gloves for gesture recognition

Area of Science:

• Human-Computer Interaction
• Robotics
• Wearable Technology

Background:

• Smart gloves are crucial for applications like entertainment, manufacturing, and rehabilitation.
• Current smart glove design is hindered by complexity, high costs, and iterative trial-and-error processes.
• Existing design platforms often require precise motion data, limiting accessibility.

Purpose of the Study:

• To present an open simulation platform for efficient smart glove design.
• To optimize sensor placement and develop deep learning models for gesture recognition.
• To reduce the cost and manual effort associated with smart glove development.

Main Methods:

• 3D hand pose extraction from monocular videos.
• Inverse kinematics for converting poses to joint angles.
• Sensor placement optimization using hand joint analysis.
• Deep learning model training with simulated sensor data.

Main Results:

• The platform accepts readily available video input (smartphones, webcams), unlike existing methods requiring precise motion data.
• Novel approaches minimize errors from imprecise motion extraction.
• Efficient sensor placement selection is enabled.
• A case study demonstrates a sensible smart glove design and evaluates component performance.

Conclusions:

• The proposed open simulation platform streamlines smart glove design.
• It lowers costs and manual effort by utilizing video input and advanced algorithms.
• The platform offers a viable solution for developing advanced smart glove applications.