Application of Wearable Insole Sensors in In-Place Running: Estimating Lower Limb Load Using Machine Learning.

Area of Science:

• Biomechanics
• Sports Science
• Machine Learning

Background:

• High-intensity activities cause musculoskeletal injuries, particularly tibia fatigue damage.
• Current monitoring methods lack synchronized multi-source data and deep signal analysis.
• Wearable algorithms often miss crucial deep signal features for early injury detection.

Purpose of the Study:

• To develop a machine learning model for estimating vertical ground reaction force (vGRF) and tibia bone force (TBF) using insole pressure signals.
• To create a synchronized dataset combining laboratory equipment and wearable insole data for training and validation.
• To enhance the early detection and prevention of tibial fatigue injuries.

Main Methods:

• Simultaneous data collection from laboratory equipment and wearable insole sensors during in-place running.
• Development of a machine learning model integrating Temporal Convolutional Network (TCN) and Transformer modules.
• Introduction of a Weight-MSELoss function to improve peak prediction accuracy.

Main Results:

• Achieved a normalized root mean square error (NRMSE) of 7.33% for vGRF prediction.
• Achieved a normalized root mean square error (NRMSE) of 10.64% for TBF prediction.
• The model effectively integrates local feature extraction and global modeling for accurate force estimation.

Conclusions:

• The WearLab-Leg dataset and proposed TCN-Transformer model provide a convenient solution for biomechanical monitoring.
• Offers reliable data and technical support for early warnings of fatigue-induced injuries in athletes and patients.
• Advances the potential for non-invasive, accurate assessment of tibial load and injury risk.