Foot Strike Angle Prediction and Pattern Classification Using LoadsolTM Wearable Sensors: A Comparison of Machine...

Area of Science:

• Biomechanics
• Sports Science
• Wearable Technology

Background:

• Foot strike pattern is crucial for runners and performance analysis.
• Wearable sensors offer a versatile method for collecting running data.
• Loadsol™ pressure insoles present an opportunity for gait analysis.

Purpose of the Study:

• To predict foot strike angle and classify foot strike patterns.
• To evaluate machine learning techniques (multiple linear regression, conditional inference tree, random forest) for this task.
• To assess model performance using 3D kinematics as ground truth.

Main Methods:

• Utilized Loadsol™ wearable pressure insoles.
• Applied three machine learning techniques: multiple linear regression (MR), conditional inference tree (TREE), and random forest (FRST).
• Validated model accuracy against 3D kinematic data.

Main Results:

• Prediction model accuracy was comparable across all methods (RMSE: MR=5.16°, TREE=4.85°, FRST=3.65°).
• Random forest and regression models showed lower maximum error (14.3°, 13.75°) compared to the inference tree (19.02°).
• Classification performance exceeded 90% for all models (MR=90.4%, TREE=93.9%, FRST=94.1%).

Conclusions:

• Wearable pressure insoles and machine learning can accurately predict and classify running foot strike patterns.
• All tested machine learning models demonstrated high classification performance.
• Further training data including more mid-foot strikes may improve classification accuracy for this pattern.