Monocular Gait Analysis via Convolutional Neural Networks
Solution Overview
Problem
Current methods for assessing the risk of fall in elderly individuals rely on subjective observations of gait parameters, which can lack accuracy and lead to inappropriate treatment decisions.
Innovation Solution
A computer-implemented method and system for human gait analysis using a monocular camera to obtain a video stream, which infers three-dimensional gait information through convolutional neural networks, allowing for objective assessment of gait parameters relevant to fall risk analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If gait parameters are assessed through subjective observation by medically trained persons, then the assessment process is simple and requires no complex equipment, but the accuracy and objectivity of gait parameter measurement deteriorates
Solution Approach 1:
The patent replaces the manual mechanical observation system with an automated computer vision system using monocular cameras and convolutional neural networks. This substitution enables objective, accurate extraction of gait parameters (step length, walking speed, step height) from video streams, resolving the contradiction between measurement simplicity and measurement accuracy.
2Measurement precision
If high-precision gait analysis tools such as gait carpets or multiple camera systems are used, then the accuracy of gait parameter measurement is improved, but the device complexity and setup requirements increase
Solution Approach 1:
The patent extracts and isolates the essential function of multi-camera systems by using a single monocular camera combined with AI algorithms. This extraction approach achieves high-precision gait parameter measurement without requiring complex multi-camera setups or specialized equipment like gait carpets, thereby resolving the contradiction between measurement accuracy and device complexity.
Solution Approach 2:
The patent introduces convolutional neural networks as an intermediary that processes video streams from monocular cameras to extract accurate three-dimensional gait information. This AI intermediary enables high-precision measurement while avoiding the need for complex physical measurement systems.
3Productivity
If subjective gait assessment is used to assess fall risk, then the assessment process remains simple and quick, but the reliability of treatment decisions deteriorates due to potential inaccuracies
Solution Approach 1:
The patent replaces subjective human observation with automated computer vision and AI analysis, maintaining rapid assessment capability while significantly improving the reliability of gait parameter measurement and subsequent fall risk assessment. The system processes video streams in real-time to provide objective, accurate results.
Data Source
AI summary
A computer-implemented method, computer program product and computer system for human gait analysis are described. A video stream of an individual walking includes a plurality of frames reflecting the walk of the individual. A skeletal motion of the individual is extracted, by inferring, from the obtained video stream, three-dimensional gait information. The three-dimensional gait information includes estimates of the individual's joint locations including at least the individual's foot locations on each frame, wherein the estimates are derived by matching for each frame two-dimensional coordinates of the respective frame with respective three-dimensional model information of the individual's body. One or more gait parameters of the individual may be determined, based on the individual's foot locations in local extrema frames showing local extrema of the distance between one foot location of the individual and a corresponding reference joint location.



