Ambient Sensor Fall Detection via Body Configuration Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fall detection systems are inadequate for detecting when a person is positioned on the floor without a preceding rapid fall, as they often require wearable devices, historic trajectory data, and cannot distinguish between falling and non-falling trajectories, leading to missed cases of individuals becoming trapped on the floor due to voluntary descent or slow transitions.
Innovation Solution
The system uses ambient sensors to detect the center of mass and configuration of objects in relation to a surface, determining if an object is on-the-floor without requiring a worn device or historic trajectory data, utilizing visual, spatial, and depth data to identify objects and their proximity to surfaces, and sending alerts for potential falls or trapped situations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If wearable devices are used for fall detection, then detection accuracy for rapid falls is improved, but user compliance and accessibility deteriorate
Solution Approach 1:
The patent replaces wearable mechanical/sensor systems with an ambient optical sensing system using depth cameras and image processing. The system uses visual configuration analysis (joint positions, body posture, center of mass location) instead of mechanical accelerometers or gyroscopes, eliminating the need for wearables while maintaining detection capability for both rapid falls and slow transitions to floor position.
Solution Approach 2:
The patent introduces an intermediary computational model that processes ambient visual data to infer fall states. Instead of directly measuring physical quantities with wearables, the system uses image processing algorithms to extract body configuration features (joint locations, aspect ratios, center of mass) as intermediaries to determine whether a fall has occurred, bridging the gap between passive ambient sensing and active fall detection.
2Measurement precision
If trajectory tracking is used to detect falls, then detection of rapid falls is improved, but detection of slow or voluntary transitions to floor deteriorates
Solution Approach 1:
The patent inverts the traditional fall detection approach by not tracking the descent trajectory forward in time, but rather analyzing the final body configuration state after the movement has completed. By examining the spatial arrangement of body joints, center of mass position, and overall posture in the final frame, the system can identify floor contact regardless of how the person got there (rapid fall, slow slide, or voluntary descent).
Solution Approach 2:
The patent performs preliminary identification of body configuration parameters (joint positions, center of mass, aspect ratio) from ambient visual data before making the fall determination. This preliminary extraction of spatial features enables the system to assess the final state comprehensively, allowing detection of all descent types without requiring continuous trajectory tracking throughout the movement.
3Measurement precision
If historic trajectory data is required for fall detection, then accuracy for rapid falls is improved, but real-time detection capability and response time deteriorate
Solution Approach 1:
The patent extracts only the essential spatial configuration parameters (body joint positions, center of mass location, overall posture orientation) from the visual data at the moment of detection, discarding the need for historic trajectory information. By taking out just the critical final-state features needed for fall assessment, the system achieves rapid real-time detection without the computational burden of analyzing entire movement histories.
4Measurement precision
If the system must distinguish between falling and non-falling trajectories, then accuracy for rapid falls is improved, but detection of ambiguous cases (slow transitions, voluntary descent) deteriorates
Solution Approach 1:
The patent changes the detection parameters from dynamic trajectory characteristics (speed, acceleration, path curvature) to static spatial configuration parameters (body joint positions relative to each other, center of mass height, overall body aspect ratio, orientation angles). This parameter transformation allows the system to evaluate the final body state without needing to classify the type or speed of descent, naturally handling ambiguous cases where the distinction between falling and non-falling is unclear.
Data Source
AI summary
Presented herein are systems and methods for detecting that a human or other object is on-the-floor, e.g., after a fall or other means of descent to the floor. The disclosed technology does not require a historic trajectory of movement, nor does the disclosed technology require a worn device. Detection is performed by determining the configuration of the subject using environmental data, such as visual data or depth data.