Monocular Video Gait Characterization via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gait analysis methods rely on complex and costly equipment such as pressure-sensitive walkways and multi-camera motion capture systems, limiting their availability and accessibility for widespread use in clinical settings.
Innovation Solution
The development of video-based gait characterization and analysis systems that utilize monocular video input from a single camera to quantify gait kinematics and detect gait events, reducing hardware requirements and enabling analysis in various environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialized equipment such as pressure-sensitive walkways and multi-camera motion capture systems is used, then measurement precision of gait parameters is improved, but device complexity and cost increase
Solution Approach 1:
The patent uses a single camera to capture video images that serve as a visual copy of the subject's gait motion, replacing the need for complex multi-camera motion capture systems. The video feed is then processed through machine learning models that extract keypoint coordinates and compute gait parameters from this copied visual information, achieving accurate measurement without specialized equipment.
Solution Approach 2:
The patent replaces mechanical measurement systems (pressure-sensitive walkways, multi-camera hardware) with a computational approach using a single camera and machine learning algorithms. The mechanical complexity of multiple synchronized cameras and pressure sensors is substituted with software-based processing of video data, maintaining measurement precision while reducing hardware complexity.
2Measurement precision
If specialized equipment such as pressure-sensitive walkways and multi-camera motion capture systems is used, then measurement precision of gait parameters is improved, but cost increases
Solution Approach 1:
The patent employs a standard single camera and free or low-cost machine learning models (such as MediaPipe or OpenPose) instead of expensive specialized equipment. This approach uses inexpensive, readily available components to achieve accurate gait analysis, making the system accessible for widespread clinical use without requiring costly infrastructure.
Solution Approach 2:
The system creates a digital copy of gait motion through video imaging and processes it computationally, eliminating the need for expensive physical measurement infrastructure. This copying approach allows standard cameras to replace costly motion capture systems while maintaining measurement accuracy through sophisticated algorithmic processing.
3Measurement precision
If specialized equipment is used, then gait analysis accuracy is improved, but ease of operation decreases due to limited availability
Solution Approach 1:
The patent makes gait analysis universally accessible by using a single standard camera that can be found in most clinical settings, rather than requiring specialized motion capture laboratories. The system processes video data through machine learning models to provide comprehensive gait analysis, enabling the same functionality to be deployed across diverse environments from small clinics to large hospitals without specialized infrastructure.
Data Source
AI summary
Gait characterization and analysis based on a monocular video of a walking subject are achieved, in various embodiments, by processing the video using one or more machine-learning models to predict three-dimensional (3D) keypoint coordinates for a set of anatomical keypoints of the subject, along with joint angles, body-segment rotations, and/or gait event classifications, for multiple video frames. The machine-learning models may be trained using ground-truth data acquired with a marker-based motion capture system and/or pressure-sensitive walkway.