Real-Time Gait Event Detection Based on Kinematic Data Coupled to a Biomechanical Model

Area of Science:

• Biomechanics
• Robotics
• Neuroscience

Background:

• Real-time detection of gait events like initial contact (IC), foot flat (FF), heel off (HO), and toe off (TO) is crucial for advancing neurorobotic (NR) and neuroprosthetic (NP) control.
• Current methods may have limitations in precision or speed, impacting the effectiveness of NR/NP systems.

Purpose of the Study:

• To develop and validate real-time, threshold-based algorithms for detecting multiple key gait events.
• To assess the precision and timing delays of these algorithms for potential application in NR/NP control and gait analysis.

Main Methods:

• Developed three real-time threshold-based algorithms utilizing kinematic data and a biomechanical model.
• Validated algorithms using data from seven subjects walking at three speeds on an instrumented treadmill (558 steps).
• Established reference gait events using marker and force plate data.

Main Results:

• All algorithms demonstrated excellent precision with no false positives.
• Timing delays for initial contact (IC) and toe off (TO) detection were comparable to state-of-the-art methods.
• Significantly lower timing delays were achieved for foot flat (FF) detection.
• Heel off (HO) detection was less precise, limiting its immediate application.

Conclusions:

• The developed algorithms offer high precision and low latency for IC, FF, and TO detection, suitable for NR/NP control.
• Kinematic data's availability and minimal computational burden make these algorithms practical for real-time applications.
• These methods are also applicable for pathological gait screening and general gait analysis in diverse settings.