Benchmarking of IMU-Based Gait Event Detection Algorithms Across Diverse Terrain Conditions

Area of Science:

• Biomechanics
• Wearable Technology
• Human Locomotion Analysis

Background:

• Accurate gait event detection is crucial for analyzing human locomotion.
• Existing algorithms are primarily optimized for flat surfaces, limiting their real-world applicability.
• Ecological terrains present unique challenges for gait analysis.

Purpose of the Study:

• To evaluate the performance of various gait event detection algorithms on irregular terrains.
• To assess algorithm accuracy in temporal gait segmentation across diverse ground conditions.
• To identify robust algorithms for real-world gait analysis.

Main Methods:

• Nine healthy volunteers were equipped with 17 inertial sensors.
• Participants walked on 12 distinct terrains with varying slopes.
• Algorithm performance was benchmarked against an optoelectronic system using precision, recall, and F1-score.

Main Results:

• Algorithm performance generally decreased on irregular terrains compared to flat surfaces.
• The Paraschiv-Ionescu algorithm demonstrated superior performance, achieving near-perfect F1-scores across most conditions.
• Terrain complexity significantly impacted gait event detection accuracy.

Conclusions:

• Gait event detection algorithm performance degrades with increasing terrain complexity.
• The Paraschiv-Ionescu method shows promise for robust gait analysis on varied terrains.
• The EUROBENCH framework effectively facilitates locomotion testing in realistic conditions.