Development and Evaluation of a Slip Detection Algorithm for Walking on Level and Inclined Ice Surfaces

Area of Science:

• Biomechanics
• Human Factors Engineering
• Sports Medicine

Background:

• Slip-resistant footwear is crucial for preventing fall injuries on icy surfaces.
• The Maximum Achievable Angle (MAA) test assesses footwear slip resistance but relies on subjective human observation, introducing variability.
• Automated slip detection is needed to enhance the objectivity and reliability of the MAA test.

Purpose of the Study:

• To develop and evaluate an automated slip detection algorithm for analyzing walking on level and inclined ice surfaces.
• To integrate this algorithm into the MAA test protocol, replacing the need for human observers.
• To improve the precision and reduce variability in assessing winter footwear slip resistance.

Main Methods:

• Collected kinematic data using an optical motion capture system from participants walking on ice surfaces (0-12° incline).
• Developed an algorithm to segment data into steps and extract features for machine learning classifiers.
• Utilized two linear support vector machine classifiers trained to differentiate between toe and heel slips.

Main Results:

• The algorithm achieved an overall F1 score of 90.1% for slip detection across approximately 11,000 steps, including ~4700 slips.
• High F1 scores were reported for classifying specific slip types: backward toe slips (97.3%), backward heel slips (80.9%), and forward heel slips (86.5%).
• Forward toe slip classification achieved an F1 score of 54.5%.

Conclusions:

• The developed automated slip detection algorithm effectively identifies slips during ice walking, offering a more objective measure than human observation.
• This algorithm has the potential to significantly enhance the reliability and accuracy of the MAA test for evaluating winter footwear.
• Further refinement may be needed to improve the classification accuracy of specific slip types, such as forward toe slips.