A machine learning approach to estimate Minimum Toe Clearance using Inertial Measurement Units.

Area of Science:

• Biomechanics
• Wearable Technology
• Machine Learning

Background:

• Falls are a leading cause of injury and death in older adults, with tripping during walking being a major contributor.
• Minimum toe clearance (MTC) is a critical gait parameter for assessing tripping risk.
• Wearable Inertial Measurement Units (IMUs) offer a practical solution for gait monitoring, but noise limits accuracy in estimating MTC height.

Purpose of the Study:

• To develop and validate a machine learning approach for accurate MTC height estimation using IMU data.
• To investigate the effectiveness of Generalized Regression Neural Networks (GRNN) with selected features for MTC height prediction.
• To assess the potential for real-time, out-of-laboratory gait monitoring for fall prevention.

Main Methods:

• Utilized Inertial Measurement Units (IMUs) containing accelerometers and gyroscopes to collect gait data.
• Employed a machine learning approach, specifically Generalized Regression Neural Networks (GRNN), trained on raw and integrated inertial signals.
• Implemented a hill-climbing feature-selection method to identify optimal features for MTC height estimation.

Main Results:

• The GRNN model achieved a root-mean-square-error (RMSE) of 6.6mm for young adults and 7.1mm for older adults.
• The developed method demonstrated approximately 68% less RMSE compared to existing MTC height estimation techniques.
• The study identified optimal feature sets (9 for young adults, 5 for older adults) for accurate prediction.

Conclusions:

• The proposed GRNN-based MTC height estimation method shows high accuracy and significant improvement over previous techniques.
• This approach holds strong potential for real-time gait monitoring during everyday activities, aiding in the prevention of falls in older adults.
• Practical, wearable gait analysis using IMUs can provide valuable insights into MTC height for enhanced fall risk assessment.