Error distribution and post hoc disturbance features in IMU-based hip joint angle estimation during complex movements

Area of Science:

• Biomechanics
• Wearable Technology
• Human Movement Analysis

Background:

• Current evaluations of IMU-based joint angle estimation often use global metrics, potentially masking errors during complex movements.
• Understanding localized errors is crucial for improving the accuracy of wearable motion tracking systems.

Purpose of the Study:

• To analyze the local error distribution of IMU-based left hip angle estimation during various movement categories.
• To investigate the relationship between estimation errors and specific movement phases and segments.
• To examine how post hoc disturbance features correlate with error patterns.

Main Methods:

• Analysis of the Hip-ROM-Y dataset from 16 healthy male participants.
• Utilized leave-one-subject-out validation for a unified left hip joint predictor.
• Characterized errors by movement category, normalized movement phase, and high-error segments.

Main Results:

• Complex movements generally yielded higher estimation errors compared to basic movements.
• Error increases were not uniform, primarily occurring in mid-phase segments of complex movements.
• Increased errors correlated with greater deviations in the pelvis-thigh segment's relative orientation.

Conclusions:

• Evaluations of IMU-based hip angle estimation should consider local error distributions, not just global metrics.
• Focusing on local errors during complex movements provides a more nuanced understanding of wearable sensor performance.
• The stability of the pelvis-thigh segment relationship is a key factor influencing hip angle estimation accuracy.