A Low-Cost Wearable Hand Gesture Detecting System Based on IMU and Convolutional Neural Network

Area of Science:

• Biomedical Engineering
• Robotics
• Machine Learning

Background:

• Accurate hand gesture recognition is crucial for advanced human-computer interaction and rehabilitation.
• Existing systems can be costly or lack comprehensive kinematic data capture.

Purpose of the Study:

• To develop a low-cost, wearable system for detecting hand gestures using distributed inertial measurement units (IMUs).
• To enable wireless transmission of hand kinematic data for remote processing.
• To classify gestures and determine finger spatial locations using a convolutional neural network (CNN) and modified Denavit-Hartenberg notation.

Main Methods:

• A distributed multi-node system with IMUs and a central microcontroller was designed.
• Hand kinematic information was captured and transmitted wirelessly.
• A CNN model was employed for gesture recognition and classification.
• Modified Denavit-Hartenberg notation was utilized for finger spatial localization.

Main Results:

• The system successfully recognized various hand gestures.
• It accurately captured and displayed the orientation and posture of individual fingers.
• The prototype demonstrated effective hand kinematic information acquisition.

Conclusions:

• The developed low-cost wearable system offers a viable solution for hand gesture detection.
• The system has potential applications in hand rehabilitation evaluation and human-computer interaction.
• Integration of IMUs, CNNs, and Denavit-Hartenberg notation provides comprehensive hand kinematics analysis.