Skip to contents

Detect and correct OpenPose gait anomalies

Source: R/posefixer.R
posefixer.Rd

The main PoseFixeR workflow. Given imported and reshaped OpenPose keypoint trajectories, it normalises them, detects and corrects left/right leg swaps, anatomical (segment-length), biomechanical (joint-angle / range-of-motion) and physical (frame-transition) anomalies as well as low-confidence detections, sets anomalous keypoints to NA, standardises segment lengths, then imputes and smooths the trajectories.

Usage

posefixer(dat, race, fps)

Arguments

dat

Imported and reshaped OpenPose data (see import_json() and reshaping()): a list with x, y and confidence keypoint matrices (frames x keypoints).

race

Reference-population index for the anatomical proportions (1 or 2).

fps

Frames per second of the source video.

Value

The corrected data in the same list structure as dat, or NA if the recording fails the quality criteria.

References

Sugiyama Y, Uno K, Matsui Y (2023). Types of anomalies in two-dimensional video-based gait analysis in uncontrolled environments. PLOS Computational Biology 19:e1009989. doi:10.1371/journal.pcbi.1009989