Human Joint Angle Estimation with Inertial Sensors and Validation with A Robot Arm

Area of Science:

• Biomechanics
• Sensor Technology
• Robotics

Background:

• Traditional motion capture systems are expensive, require controlled environments, and are prone to occlusion.
• Wearable inertial sensors (accelerometers, gyroscopes, magnetometers) offer a low-cost, occludable-free alternative for human movement tracking.
• Accurate human joint angle estimation is crucial for various applications.

Purpose of the Study:

• To report significant improvements to a previously developed kinematic arm model and Unscented Kalman Filter (UKF) algorithm for human joint angle tracking.
• To enhance the algorithm by incorporating sensor drift models, range-of-motion constraints, and zero-velocity updates.
• To rigorously assess the performance of the improved algorithm using a high-precision industrial robot arm.

Main Methods:

• Developed an improved human joint angle tracking algorithm based on a kinematic arm model and the Unscented Kalman Filter (UKF).
• Incorporated gyroscope and accelerometer random drift models, physical joint range-of-motion constraints, and zero-velocity updates.
• Validated the algorithm's performance against a high-precision industrial robot arm over extended recording durations and varying movement speeds.

Main Results:

• The enhanced algorithm demonstrated excellent agreement with the robot arm reference, achieving an average Root Mean Square (RMS) angle error of approximately 3 degrees for all six joint angles.
• The Unscented Kalman Filter (UKF) slightly outperformed the Extended Kalman Filter (EKF).
• The system proved effective across slow, normal, and fast movements, maintaining accuracy over 15-minute recording periods.

Conclusions:

• The improved wearable inertial sensor-based algorithm provides highly accurate human joint angle tracking.
• The method overcomes limitations of traditional motion capture, offering a versatile and cost-effective solution.
• Further development of sensor fusion algorithms with drift compensation and physical constraints enhances tracking precision.