Gait Event Detection in Real-World Environment for Long-Term Applications: Incorporating Domain Knowledge Into...

Area of Science:

• Biomechanics
• Wearable technology
• Signal processing

Background:

• Gait event detection is crucial for continuous monitoring and long-term gait analysis.
• Current algorithms often rely on controlled indoor data, limiting real-world applicability.
• Real-world gait analysis requires algorithms robust to variations in speed, terrain, inclination, and turns.

Purpose of the Study:

• To develop a robust gait event detection algorithm for real-world applications.
• To incorporate domain knowledge of human gait into time-frequency analysis.
• To validate the algorithm's performance in diverse indoor and outdoor environments.

Main Methods:

• Utilized domain knowledge of human gait principles.
• Applied time-frequency analysis to long-term accelerometer signals.
• Validated the algorithm with approximately 93,600 gait events across indoor and outdoor settings.

Main Results:

• The proposed algorithm demonstrated high accuracy and robustness in detecting gait events.
• Consistent high performance was observed across all tested datasets, including both indoor and outdoor environments.
• The approach effectively addresses challenges posed by natural human locomotion.

Conclusions:

• Incorporating domain knowledge into time-frequency analysis enhances gait event detection.
• The developed algorithm is suitable for real-world gait analysis applications.
• This method offers a reliable solution for continuous and long-term gait monitoring.