Motion-Driven Automatic IMU Orientation Calibration via SO(3) Pattern Alignment

Area of Science:

• Biomedical Engineering
• Sensor Technology
• Human Motion Analysis

Background:

• Body-worn inertial measurement units (IMUs) are crucial for motion estimation.
• Sensor orientation drift in IMUs leads to inaccurate motion tracking.
• Current calibration methods often require dedicated steps and may be cumbersome.

Purpose of the Study:

• To develop a novel technique for automatic calibration of body-worn IMUs.
• To recover IMU alignment with the body segment during normal motion.
• To eliminate the need for dedicated calibration procedures.

Main Methods:

• Formulated automatic calibration as aligning motion-induced orientation distributions on SO(3).
• Employed a correspondence-free spherical pattern matching method (SO3_SPMC) for alignment.
• Utilized transformed basis vector distributions and spherical cross-correlation.

Main Results:

• Demonstrated calibration accuracy comparable to supervised single-frame calibration.
• Achieved accurate calibration without requiring explicit calibration poses or user intervention.
• Identified walking, typing, and using a computer mouse as effective activities for wrist-mounted sensor calibration.

Conclusions:

• The proposed automatic calibration framework effectively recovers IMU alignment during activities of daily living.
• This method offers a practical solution for maintaining IMU accuracy without interrupting user activity.
• Automatic calibration enhances the reliability of IMU-based motion estimation in real-world applications.