IMU-based Real-Time Crutch Gait Phase and Step Detections in Lower-Limb Exoskeletons

Anis R. Shakkour, David Hexner, Yehuda Bitton, Avishai Sintov · 2026

Read IMU-based Real-Time Crutch Gait Phase and Step Detections in Lower-Limb Exoskeletons by Anis R. Shakkour, David Hexner, Yehuda Bitton, Avishai Sintov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-10832.txt.

Introduction

Anis R. Shakkour, David Hexner, Yehuda Bitton, and Avishai Sintov build a lean gait detector for lower-limb exoskeletons. Such devices must know the user's gait phase and each step in real time. A wrong guess can fight the user's motion and risk safety. Common methods lean on complex force sensors, which add delay to the control loop.

Their framework uses a single low-cost inertial measurement unit set into the crutch hand grip. No mechanical changes to the device are needed. One small sensor, read well, replaces a rack of hardware. The work targets cheaper, safer exoskeletons and prostheses that respond in step with the people who wear them.

Catalog record

CREATORAnis R. Shakkour, David Hexner, Yehuda Bitton, Avishai Sintov
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-10832

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

IMU-based Real-Time Crutch Gait Phase and Step Detections in Lower-Limb Exoskeletons Anis R. Shakkour, David Hexner, Yehuda Bitton, Avishai Sintov Abstract Lower limb exoskeletons and prostheses require precise, real time gait phase and step detections to ensure synchronized motion and user safety. Conventional methods often rely on complex force sensing hardware that introduces control latency. This paper presents a minimalist framework utilizing a single, low cost Inertial-Measurement Unit (IMU) integrated into the crutch hand grip, eliminating the need for mechanical modifications. We propose a five phase classification system, including standard gait phases and a non locomotor auxiliary state, to prevent undesired motion. Three deep learning architectures were benchmarked on both a PC and an embedded system. To improve performance under data constrained conditions, models were augmented with a Finite State Machine (FSM) to enforce biomechanical consistency. The Temporal Convolutional Network (TCN) emerged as the superior architecture, yielding the highest success rates and lowest latency. Notably, the model generalized to a paralyzed user despite being trained exclusively on healthy participants. Achieving a 94% success rate in detecting crutch steps, this system provides a high performance, cost effective solution for real time exoskeleton control. I Introduction Lower-limb exoskeletons are transformative robotic devices that either assist in rehabilitation [ 8 ] or restore standing and walking ability to individuals with severe mobility impairments, such as those resulting from spinal cord injury [ 2 ] . Similarly, active lower-limb prostheses enable amputees to restore mobility and achieve a more natural gait [ 1 ] . These assistive devices are fundamentally designed to enhance user safety, promote smooth and natural motion, and maximize comfort. However, optimal performance is unattainable without a robust and functional gait phase detection method [ 20 ] . This detection unit is essential, as it provides the device’s main controller with real-time information about the user’s current gait phase. By accurately detecting the current phase (e.g., heel strike, swing), the device can produce the corresponding, temporally synchronized motion and actuation functionality. Consequently, integrating a highly accurate and low-latency gait detection strategy is paramount for improving the overall performance, responsiveness, and safety of the control system in the walking device. Fig. 1: A participant walking with the ReWalk [ 6 ] lower limb exoskeleton and crutches. The right-hand side crutch is equipped with an Inertial Measurement Unit (IMU) for gait phase detection. Current methods for gait phase detection typically employ various sensor modalities. The detection strategies can be clearly separated based on the location of the sensing hardware. Sensors mounted directly on the robotic device commonly include joint encoders and torque sensors to measure the robot’s state, and force sensors integrated into the foot sole to detect ground contact [ 3 , 18 ] . While these provide direct kinematic and kinetic feedback, they primarily measure the current device state and do not interact with the user to capture intent [ 2 ] . This reliance on hindsight data precludes intent recognition, as the system only observes motion after sensing a shift in the body center of mass. Such reactive sensing fails to detect intent, resulting in sluggish coordination that lacks the proactive responsiveness required for natural walking. Furthermore, relying solely on foot sensors can lead to inconsistent performance across varied terrains or for users with atypical gait patterns. A comprehensive, yet challenging, approach involves Electro-Myography (EMG), which detects muscle activation to predict intent, but its real-world utility is limited by electrode placement sensitivity, signal noise, and inter-subject variability [ 11 , 9 ] . An alternative and often more proactive approach utilizes sensors placed on the assistive devices, primarily the crutches, which the user manipulates to express walking intent [ 12 , 13 ] . In addition to assisting in balance, the crutch acts as the primary mechanical interface between the user’s intention and the ground. Instrumented crutches typically incorporate force sensors to directly quantify the axial load and shear forces exerted by the user onto the crutch, specifically measuring the interaction forces between the crutch tip and the ground [ 4 , 7 ] . An advanced and highly functional approach integrates these kinetic measurements with Inertial Measurement Units (IMUs) [ 16 , 15 ] . This sensor fusion provides a comprehensive dataset for accurately modeling the crutch’s state, providing a crucial predictive window for the exoskeleton controller [ 10 ] . Crucially, while IMUs are easy to deploy, the viability of relying solely on the crutch’s inertial data, without any supporting force sensors, for robust, multi-phase gait classification has not yet been thoroughly explored, presenting a significant opportunity for simplification and cost reduction. This gap motivates our work. In this paper, we explore the use of an IMU on a crutch for cost-effective gait phase and step detections. We present a novel, minimalist, and computationally efficient framework for crutch gait detection tailored for real-time exoskeleton control (Figure 1 ). Our key contribution lies in demonstrating that gait detection can be achieved using signals from a single, low-cost and standard IMU concealed within on the crutch hand grip, completely eliminating the need for force sensors or any mechanical modification to the crutch tip. While the common approach for control systems is to classify the standard four phases of the crutch gait cycle (i.e., stance, take-off, swing and strike [ 5 ] ), a more robust controller requires the addition of a crucial functional phase. We extend the model to a five-phase system by adding the Auxiliary phase covering any non-gait activities such as standing, drinking, scratching or opening a door. This inclusion is vital for safety, as this phase allows the control system to enter a locked state, preventing undesired motion caused by incidental sensor readings during non-locomotor activities. We consider a temporal-based classification model to achieve accurate and low-latency performance. The resulting methodology offers a high-performance and cost-effective solution to significantly improve human-exoskeleton coordination and responsiveness. II Method II-A System In this work, the crutch system is the dedicated interface for the detection of user gait intent, utilizing a standard forearm crutch instrumented with a single, low-cost off-the-shelf IMU. The IMU was integrated within the hand grip of the right hand side crutch as seen in Figure 2 . Specifically, the z -axis is oriented vertically upward, aligned with gravity during standing, the x -axis is pointed horizontally toward the walking direction and the y -axis is pointed perpendicular to the user’s walking direction. This specific alignment ensures that the output signals, such as forward pitch rotation about the x -axis, directly and naturally correspond to key crutch motions like the swing phase. Fig. 2: The crutch with an integrated IMU within the hand grip. The IMU sensor provides streaming at 100 Hz including all nine degrees of freedom: tri-axial linear acceleration ( a_{x} , a_{y} , a_{z} ), tri-axial angular velocity ( \omega_{x} , \omega_{y} , \omega_{z} ) and tri-axial magnetic field strength ( \mu_{x} , \mu_{y} , \mu_{z} ). Additionally, the IMU’s internal sensor fusion engine provides pre-processed orientation outputs in both Euler angles and quaternions, offering flexibility in choosing input features for the classification model. Consequently, a processed measurement vector at time t is defined by \mathbf{x}_{t}\in\mathbb{R}^{9} . This vector is composed of the tri-axial raw

Read the complete text