An adaptive, real-time cadence algorithm for unconstrained sensor placement.

Area of Science:

• Biomechanics
• Wearable Technology
• Signal Processing

Background:

• Conventional cadence detection algorithms struggle with unconstrained sensor placement, leading to inaccurate step or stride identification.
• Sensor position and orientation significantly impact the reliability of existing correlation-based methods, causing false positives and negatives.

Purpose of the Study:

• To evaluate a novel adaptive real-time cadence detection algorithm (CDA) for unconstrained sensor placement.
• To compare the performance of the CDA, utilizing Sylvester's criterion (SC) for stride validation, against conventional correlation procedures.

Main Methods:

• The CDA was tested with 22 volunteers performing walking, running, and sprinting across various circuits and gait-related activities.
• The algorithm was evaluated across 10 different sensor placements, with reference strides determined by an offline foot sensor algorithm.
• Sylvester's criterion was employed to validate correlation peaks as strides, enhancing accuracy.

Main Results:

• The CDA demonstrated high accuracy with true positive rates of 85.6-100.0% and low false positive rates of 0.0-2.9%.
• Consistent cadence feedback accuracy of 98.9% ± 0.2% was achieved across all circuits, subjects, and sensor positions.
• The CDA significantly outperformed conventional cross-correlation (87.3% ± 13.5%) and autocorrelation methods (73.0-82.2%).

Conclusions:

• Sylvester's criterion significantly improves the accuracy and robustness of cadence detection.
• The developed CDA provides reliable cadence feedback, even with unconstrained sensor placement and varied gait activities.
• This algorithm offers a superior solution for real-time cadence monitoring in diverse applications.