Monocular Skeletal Pose Inference for Privacy-Safe Fall Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fall detection systems, such as wearable devices and remote monitoring services, are costly and do not address the need for scalable, cost-effective solutions that can determine a subject's pose and respond to adverse events, such as falls, while maintaining privacy and compliance with regulations.
Innovation Solution
A monocular camera-based system that uses a pose estimation model to identify key points on a subject, classify their pose, and send alerts to client devices, while obscuring identity to ensure compliance and privacy, using a cloud-based system for centralized monitoring across healthcare facilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If expensive 3D time of flight cameras or stereo vision systems are used for pose detection, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces expensive, complex 3D time of flight cameras and stereo vision systems with inexpensive monocular RGB cameras. This substitution uses readily available, low-cost hardware while achieving comparable pose detection accuracy through advanced image processing algorithms that extract skeletal information from 2D images.
Solution Approach 2:
The patent replaces complex optical-mechanical systems (stereo cameras, time of flight sensors) with a simpler monocular camera system combined with computational algorithms. The mechanical/optical complexity is substituted with software-based pose estimation that processes 2D images to infer 3D skeletal positions.
2Reliability
If traditional fall detection systems are used, then reliability is improved, but cost increases
Solution Approach 1:
The patent employs budget-friendly monocular cameras instead of expensive specialized fall detection hardware. The system achieves reliable fall detection through cost-effective RGB cameras combined with intelligent image processing that analyzes pose changes to identify falls.
Solution Approach 2:
The system uses the monocular camera to perform multiple functions - not only capturing images but also automatically detecting poses, identifying falls, and triggering alerts without requiring additional specialized sensors or devices. The single camera serves the entire fall detection function.
3Productivity
If real-time monitoring is implemented, then productivity is improved, but loss of information (privacy) increases
Solution Approach 1:
The patent extracts only the essential information needed for fall detection - skeletal pose data and fall events - while discarding or obscuring identifiable personal information. The system processes images to extract pose keypoints and skeletal structures, then deliberately removes or blurs facial features and other identifying characteristics before storage or transmission.
Solution Approach 2:
The patent introduces pose estimation algorithms as an intermediary layer between the camera and the monitoring system. This intermediary processes raw images into abstracted skeletal representations, enabling fall detection while naturally obscuring personal identity. The skeletal pose data serves as a mediator that preserves monitoring functionality while protecting privacy.
Data Source
AI summary
Systems disclosed herein are directed to a system including at least one processing unit, and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, wherein the instructions, when executed by the at least one processing unit, cause the system to perform actions including identifying, by a processing circuitry, a subject in a room on a camera feed received from a monocular camera via a network with an objection detection model; mapping, by the processing circuitry, key points of the subject in the camera feed with a pose estimation model; classifying, by the processing circuitry, a pose of the subject based on the key.