Related Experiment Video
Updated: May 12, 2026
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.7K
Enhancing fall risk assessment: instrumenting vision with deep learning during walks.
Jason Moore1, Robert Catena2, Lisa Fournier2
1Department of Computer and Information Sciences, Northumbria University, Newcastle, NE1 8ST, UK.
Journal of Neuroengineering and Rehabilitation
|June 22, 2024
Summary
This study introduces VARFA, a deep learning algorithm that automatically analyzes video data to assess visual attention during walking. This objective approach enhances fall risk assessment by providing crucial behavioral and contextual data.
Area of Science:
- Biomechanics and Movement Science
- Computer Vision and Machine Learning
- Clinical Assessment and Rehabilitation
Background:
- Falls are a significant clinical concern, with current risk assessments relying heavily on subjective visual observation of gait.
- Subtle gait deficits, crucial for fall risk, are often missed by traditional observational methods.
- Objective measures like inertial measurement units (IMUs) capture gait characteristics but lack environmental and behavioral context.
Purpose of the Study:
- To develop an automated method for analyzing visual attention and environmental interactions during gait to improve fall risk assessment.
- To complement existing gait analysis techniques (e.g., IMUs) with objective visual and behavioral data.
- To address the limitations of manual video analysis, which is time-consuming and subjective.
Main Methods:
- Development of a deep learning-based object detection algorithm named VARFA (Vision and Real-time Fall risk Assessment).
- Training VARFA using a novel lab-based dataset with a YoloV8 model for identifying environmental objects and obstacles.
- Implementation of a U-NET based model for accurate segmentation of walking paths.
Main Results:
- VARFA achieved high accuracy (0.93 mAP50) in identifying and localizing static objects, with 93% average accuracy.
- The path segmentation model demonstrated strong performance with an IoU of 0.82, indicating precise alignment with actual walking paths.
- Both models operated in real-time, highlighting their efficiency for practical applications.
Conclusions:
- The automated vision-instrumentation approach significantly enhances the efficiency and accuracy of fall risk assessment.
- VARFA provides objective data on visual attention and environmental context, complementing IMU-based gait analysis.
- This technology holds potential for personalized rehabilitation strategies in clinical populations at high risk of falls.
