Accurate gait events detection is imperative for reliable assessment of normal and pathological gaits. However, this detection becomes challenging in the absence of force plates. Hence, this research introduces two geometric models integrated into an automatic algorithm (O-GEST) for overground gait events detection using kinematic data.

O-GEST employs B-Spline-based geometric models to represent the horizontal trajectory of foot landmarks. It leverages gait-dependent thresholds, along with optimal coefficients to detect events and compute spatiotemporal parameters on healthy and pathological gaits. To validate the proposed algorithm, timing differences in the detected events using the force plates and O-GEST were calculated and also compared between different methods on the gait data of 390 subjects. This dataset includes 200 healthy subjects, 100 subjects with unilateral hip osteoarthritis, 50 stroke survivors, 26 individuals diagnosed with Parkinson’s disease, and 14 children with cerebral palsy.

The validation results show that O-GEST detects gait events with an overall accuracy of 13.5 ms for foot-strike and 12.6 ms for foot-off. It also demonstrates significantly more accurate results than the common deep learning-based and kinematic-based methods.

O-GEST offers several advantages, including its applicability for events detection across various pathologies, capability to handle noisy trajectories, and usability in the absence of certain foot landmarks. Development of such algorithms could lead to enhanced accuracy and reliability of gait analysis in force-plate-less environments, especially in markerless gait analysis setups.