Calibration of an instrumented treadmill using a precision-controlled device with artificial neural network-based...

Area of Science:

• Biomechanics
• Instrumentation
• Biomedical Engineering

Background:

• Instrumented treadmills (ITs) are crucial for gait and balance analysis, measuring reaction forces (RF) and center of pressure (COP).
• Accurate RF and COP measurements depend on regular in situ calibration of ITs to ensure reliable data and identify recalibration needs.

Purpose of the Study:

• To develop and validate a portable, precision-controlled calibration device for in situ IT calibration.
• To implement an artificial neural network (ANN)-based correction method to enhance IT measurement accuracy.

Main Methods:

• A novel calibration device applied static and dynamic loads to an IT at multiple grid-points and belt speeds.
• A three-layered, back-propagation ANN model was trained and evaluated using applied and measured RF and COP data.
• Calibration was performed without requiring a preset template, allowing for flexible application.

Main Results:

• Static calibration reduced percent errors in Fz from -1.15% to 0.02% and Px/Py errors from -1.64/-0.73 mm to 0.02/0.03 mm.
• Dynamic calibration decreased percent errors in Fz from -3.65% to 0.30% and Px/Py errors from 2.58/-4.92 mm to -0.14/-0.47 mm.
• The ANN-based method demonstrated significant error reduction for both static and dynamic loading conditions.

Conclusions:

• The developed portable calibration device and ANN correction method effectively enhance the accuracy of instrumented treadmills.
• This approach offers a practical solution for in situ calibration, improving the reliability of gait and balance assessment data.
• The findings suggest the system's utility in correcting measurement errors for vertical loads and COP in ITs.