A Wearable Gait Monitoring System for 17 Gait Parameters Based on Computer Vision

TL;DR

This work presents a cost-effective shoe-mounted gait monitoring system that uses a stereo camera to estimate 17 spatial, temporal, and spatiotemporal gait parameters while employing an FSR-based heel trigger for timing. The hardware comprises dual wearables with Raspberry Pi controllers, wide-field stereo cameras, a marker-based detection pipeline (YOLO), and triangulation to derive 3D marker positions, achieving high accuracy (>93.61%) and low drift (4.89%) over long walks. A Transformer-based person-identification experiment on the collected gait data achieved 95.7% accuracy, indicating the potential to integrate long-sequence gait data with LLMs for disease diagnosis and personalized assessment. The system is designed for real-world use, offering rapid stereo capture, low power requirements, and ease of installation, making it suitable for hospital and home monitoring and enabling richer gait datasets beyond traditional gait mats.

Abstract

We developed a shoe-mounted gait monitoring system capable of tracking up to 17 gait parameters, including gait length, step time, stride velocity, and others. The system employs a stereo camera mounted on one shoe to track a marker placed on the opposite shoe, enabling the estimation of spatial gait parameters. Additionally, a Force Sensitive Resistor (FSR) affixed to the heel of the shoe, combined with a custom-designed algorithm, is utilized to measure temporal gait parameters. Through testing on multiple participants and comparison with the gait mat, the proposed gait monitoring system exhibited notable performance, with the accuracy of all measured gait parameters exceeding 93.61%. The system also demonstrated a low drift of 4.89% during long-distance walking. A gait identification task conducted on participants using a trained Transformer model achieved 95.7% accuracy on the dataset collected by the proposed system, demonstrating that our hardware has the potential to collect long-sequence gait data suitable for integration with current Large Language Models (LLMs). The system is cost-effective, user-friendly, and well-suited for real-life measurements.

Figures (14)

• Figure 1: (a) Image of the gait monitoring system with all the components, including Raspberry Pis, batteries, FSR, stereo cameras, and PCB; (b) A volunteer walking while wearing the system attached to the side of the shoe, with a custom-designed marker placed on the tip.
• Figure 2: Diagram illustrating the overall operation of the gait monitoring system. The large FoV cameras capture a stereo image and record a timestamp each time a heel strike occurs. Subsequently, marker position extraction and stereo triangulation, along with the timestamp, are used to calculate gait parameters.
• Figure 3: (a) Stereo camera calibration setup; the center of the image depicts the holder for the stereo camera. (b) Discrepancies comparison of the x-axis focal length calibration results for the stereo camera with its holder during three months of use, relative to its initial calibration.
• Figure 4: (a) An image showing a participant's foot in mid-air during stereo imaging with the FSR placed near the foot tip; (b) Stereo imaging is obstructed when another participant wears long pants during testing.
• Figure 5: Diagram illustrating the walking phases of the right foot and the relationship between temporal parameters. The gait cycle consists of two phases, standing/swing phases, which can be further decomposed into single support and double support time (Free characters from Blender Studio).