Fall Detection Sub-Group Decision Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automatic fall detection systems in Personal Emergency Response Systems (PERS) face challenges in accurately distinguishing between falls and everyday movements, leading to a high false alarm rate, which can be detrimental as they often require manual cancellation and may not function correctly in unconscious or panicked states.
Innovation Solution
A computer-based method that dynamically determines different decision values for sub-groups of subjects based on specific properties or characteristics, such as location, activity, or medical conditions, to improve fall detection accuracy while maintaining a low false alarm rate, by partitioning the monitored population into sub-groups and adjusting decision values accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If fall detection systems are configured to have a high false detection rate to minimize missed falls, then the true positive detection rate is improved, but the false alarm rate increases
Solution Approach 1:
The monitored population is segmented into different sub-groups based on specific properties or characteristics (e.g., location, activity, medical conditions). Each sub-group is assigned a customized decision value threshold tailored to its specific risk profile and false alarm characteristics, allowing the system to optimize detection sensitivity for each group while maintaining overall low false alarm rates
Solution Approach 2:
Different decision value thresholds are applied to different sub-groups based on their local characteristics. High-risk groups receive more sensitive thresholds (lower decision values) to maximize fall detection, while low-risk groups receive less sensitive thresholds (higher decision values) to minimize false alarms, achieving locally optimized detection quality
2Ease of operation
If a single decision value is used for all subjects, then the system is simple to operate, but it cannot optimize detection accuracy for different subject groups
Solution Approach 1:
The system automatically assigns appropriate sub-groups and decision value thresholds to subjects based on their properties and characteristics without requiring manual configuration. The automated classification and threshold assignment maintains ease of operation while achieving optimized detection accuracy for each subject group
Data Source
AI summary
Proposed are concepts for distinguishing between fall events and non-fall-events for different sub-groups within a monitored group (i.e., monitored population) of subjects. It is proposed that an entire/group population of monitored subjects may be portioned into sub-groups, each sub-group consisting of a plurality of members (i.e., subjects) having a certain property value or characteristic unique to that group. A respective decision value may be determine for each sub-group, wherein the decision value for a sub-group takes account of a previously obtained false fall detection rate for that sub-group.