Improving Data Glove Accuracy and Usability Using a Neural Network When Measuring Finger Joint Range of Motion

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Machine Learning in Healthcare

Background:

• Data gloves measure finger joint kinematics for objective hand assessment and rehabilitation.
• Existing data gloves require complex calibration, limiting use for patients with disabilities.
• Sensor stability is affected by glove construction, hand size, and material elasticity.

Purpose of the Study:

• To present a unique calibration-free approach for data glove angular calculation using a neural network.
• To improve the repeatability and accuracy of data glove measurements.
• To enable data glove use in clinical settings for individuals with limited joint mobility.

Main Methods:

• Development of a novel neural network model for angular calculation.
• Implementation of a calibration-free system for data glove operation.
• Testing and validation of the system's performance with diverse user groups.

Main Results:

• The neural network approach significantly improved data glove measurement accuracy and repeatability.
• The system demonstrated effectiveness without requiring traditional data glove calibration.
• The method is particularly beneficial for users unable to perform standard calibration procedures.

Conclusions:

• A calibration-free neural network method enhances data glove utility in clinical hand assessment and rehabilitation.
• This innovation expands the accessibility of data glove technology for patients with physical limitations.
• The findings support the integration of AI in developing more adaptable and user-friendly assistive devices.