Continuous Gait Phase Estimation for Multi-Locomotion Tasks Using Ground Reaction Force Data.

Area of Science:

• Biomechanics
• Wearable Technology
• Machine Learning

Background:

• Gait phase estimation is crucial for understanding human locomotion.
• Existing methods often rely on inertial measurement units (IMUs) and are limited to specific walking conditions.
• A need exists for robust gait analysis across varied environments.

Purpose of the Study:

• To develop and validate a real-time gait phase estimation algorithm.
• To assess algorithm performance across diverse and challenging walking conditions.
• To demonstrate the algorithm's potential for practical applications.

Main Methods:

• Utilized force sensing resistors (FSRs) integrated into insoles.
• Employed a Bidirectional Long Short-Term Memory (Bi-LSTM) deep learning model.
• Conducted experiments with ten healthy adults across various walking conditions (level ground, stairs, ramps).

Main Results:

• Achieved average gait estimation accuracy exceeding 90%.
• Reported a low root mean square error (RMSE) of 0.794.
• Obtained a high R-squared (R²) score of 0.906, indicating strong model fit.
• Demonstrated robust performance across all tested walking conditions.

Conclusions:

• The proposed FSR-based Bi-LSTM algorithm offers accurate and real-time gait phase estimation.
• The algorithm shows significant potential for widespread use in various insole-based applications.
• Applications include gait analysis, assistive device control, and motor ability assessment.