Real-Time Identification of Gait Events in Impaired Subjects Using a Single-IMU Foot-Mounted Device

Abstract

Real time identification of gait events is a mandatory condition for adaptive or patient-tailored control of robotic devices during gait therapy. Despite most of the studies in the literature have reported high accuracy in the identification of gait phases for healthy subjects, most of them were not tested on impaired subjects and/or are not suitable for real-time implementations. In this paper, we evaluated the feasibility of some of the most known algorithms for identification of gait events. We propose a novel algorithm that exploits the advantages of the different approaches used for detection of gait events. We built a wearable sensor device with a single IMU placed back of the heel. Three subjects (a healthy subject, a hemiparetic and a myelopathic) worn the devices and performed an experimental protocol with overground and treadmill walking trials. Algorithms showed a high performance for healthy gait and their suitability for real-time implementations. However, none of the algorithms in the literature could maintain high accuracy during hemiparetic or myelopathic gait. Our algorithm obtained high accuracy for the three subjects: healthy (F1-score: 0.99), hemiparetic (0.97) and myelopathic (0.96). We aim to implement our proposal as part of the control loop of a robot during robotic gait therapy.