Adaptive Pressure Threshold for Fall Detection Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fall detection systems in wearable devices face challenges in accurately distinguishing between fall events and noise, particularly due to high noise levels and drift in barometric pressure signals, leading to false positives and reduced reliability in determining if a person has gotten up after a fall.
Innovation Solution
A method is introduced to estimate noise levels in pressure signals using a confidence estimate, calculated by filtering out high-frequency noise and standard deviation, which allows for a dynamic pressure threshold adjustment, enhancing the reliability of fall detection by differentiating between true falls and noise-induced events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If barometric pressure sensors are used for fall detection, then the ability to detect falls and determine if the person has gotten up is improved, but noise levels and drift in pressure signals increase leading to false positives
Solution Approach 1:
The system performs preliminary noise characterization by collecting pressure data during known non-fall periods and calculating noise metrics (standard deviation, inter-quartile range) before actual fall detection begins. This pre-established noise baseline is then used to set adaptive thresholds that account for environmental noise levels, preventing false positives while maintaining fall detection sensitivity.
Solution Approach 2:
The patent implements dynamic threshold adjustment based on real-time noise level assessment. Instead of using fixed pressure thresholds, the system continuously monitors noise metrics and adapts the fall detection threshold accordingly. This dynamic approach allows the system to maintain reliable fall detection across varying environmental conditions while filtering out noise-induced false positives.
2Device complexity
If fixed pressure thresholds are used for fall detection, then the detection process is simplified, but accuracy decreases due to inability to distinguish between true falls and noise-induced events
Solution Approach 1:
The system performs self-characterization of noise levels by automatically analyzing its own pressure sensor output during non-fall periods. It calculates noise metrics and uses these to self-adjust detection thresholds without requiring external calibration or manual intervention. This self-service approach maintains detection accuracy while avoiding the complexity of manual threshold setting procedures.
Solution Approach 2:
The patent transforms the detection approach by changing from fixed threshold parameters to dynamic, adaptive parameters. The system modifies the pressure threshold based on measured noise characteristics (standard deviation, inter-quartile range), allowing the threshold to vary with environmental conditions. This parameter adaptation significantly improves detection accuracy while the automation keeps implementation complexity manageable.
3Reliability
If noise filtering is applied to pressure signals, then false positives are reduced, but the ability to detect subtle pressure changes may be compromised
Solution Approach 1:
The system applies partial filtering by selectively removing only the noise component from pressure signals while preserving the genuine fall-related pressure changes. It uses noise metrics calculated from non-fall periods to determine the appropriate level of filtering, applying just enough noise reduction to eliminate false positives without over-filtering legitimate fall events. This balanced approach maintains detection sensitivity while reducing false positives.
Data Source
AI summary
A wearable fall-detection device has a variety of sensors, including a pressure sensor, that provide signals for sampling environmental conditions acting on the device. An average of pressure data samples is used to determine a resultant that may indicate an amount of noise in a pressure data signal, and statistical analysis of the noise and the pressure signal average may be used to determine a confidence estimate value that indicates a level of confidence in the amount of noise that a pressure signal is subject to, or includes. The confidence estimate and known fall data, such as change in pressure between a person standing and lying, can create a threshold function that may adapt according to sampled data thus providing a customizable (either statically or dynamically) threshold function for comparing sensor data against rather than comparing data with just a linear threshold function.