Robust and Accurate Capture of Human Joint Pose Using an Inertial Sensor.

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Human Motion Analysis

Background:

• Accurate human skeletal pose estimation is crucial for diagnosing and monitoring movement disorders.
• Wearable inertial measurement units (IMUs) offer a non-invasive solution for real-time motion capture.
• Current model-based approaches face challenges in dynamic limb orientation estimation, especially for patients with disabilities.

Purpose of the Study:

• To improve the accuracy and robustness of real-time human skeletal pose estimation using IMUs.
• To address the challenges of dynamic limb orientation estimation in the context of movement disabilities.
• To present a practical and wearable solution for long-term patient monitoring.

Main Methods:

• Utilized measurement conversion techniques to linearize the inherently non-linear pose estimation problem.
• Employed quaternions instead of Euler angles to prevent Gimbal lock.
• Developed an optimization-based mathematical justification for quaternion normalization.
• Configured a robust extended Kalman filter to integrate these methods for enhanced performance.

Main Results:

• The proposed approach significantly improves the estimation accuracy of limb orientation.
• The use of quaternions effectively avoids Gimbal lock issues.
• Quaternion normalization via optimization enhances system stability.
• The integrated system provides a structured and comprehensive solution for IMU-based pose estimation.

Conclusions:

• Measurement conversion, quaternions, and an optimized Kalman filter provide a pragmatic and effective method for IMU-based human pose estimation.
• This approach is particularly beneficial for capturing subtle movement patterns in patients with disabilities.
• The system offers a robust and wearable solution for real-time monitoring and assessment.