Gait analysis methods’ review, extension and validation using magnetometer-free inertial data

Abstract

Gait analysis using inertial measurements units (IMU) offers a wider range of capabilities and freedom compared to other foot motion capture techniques, such as optical markers tracking or instrumented treadmills. However, IMU-based gait analysis is still under development and its accuracy is still being reviewed against other ground truth reference systems. This thesis aims to review and make a compendium of the existing methods, create a documentation for the gaitt project, and serve as a reference guide to newcomers. Beyond this objective, among its goals are also to implement a new rest phase detection algorithm and tune one of the parameters of an orientation estimation algorithm previously developed. For these latter two, the experimental framework consists on recorded data 92 trials of subjects walking on an instrumented treadmill, and equipped with optical markers and and IMU on each foot. This provides two ground truth reference frames against which to compare the ones obtained from the IMU recorded data. After the implementation, the new rest detection method yields very similar results to the existing one, both in gait phases duration, average stride length, and average gait velocity. The tuning of the orientation estimation algorithm was posed to compute the orientation of the IMU more accurately, but actual results from it implementation have proven to be diminishing and give no reason to use it whatsoever.