Fall Detection Algorithm False Positive Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fall detection devices often generate false positive indications, struggling to distinguish between actual falls and non-fall events such as drops or sits, which can lead to unnecessary emergency responses.
Innovation Solution
A method that uses accelerometer and barometric pressure data to differentiate between falls and drops by setting free-fall thresholds based on empirical data, correlating current event data with known fall and non-fall event signatures, and adjusting for device orientation and spin, thereby reducing false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the fall detection device uses accelerometer and barometer to detect falls, then the detection capability is improved, but the number of false positive indications increases
Solution Approach 1:
The fall detection algorithm is segmented into multiple distinct phases: free-fall detection phase (detecting when acceleration magnitude drops below threshold), impact detection phase (detecting the sudden acceleration spike upon ground contact), and post-fall verification phase (checking for characteristic motion patterns after impact). This segmentation allows the system to analyze each phase separately with phase-specific criteria, improving overall detection accuracy while reducing false positives from non-fall events that may not exhibit all characteristic phases.
Solution Approach 2:
The system dynamically adjusts detection parameters based on the detected phase. During free-fall, the system looks for sustained low acceleration; during impact, it detects sudden acceleration spikes; after impact, it verifies characteristic motion patterns. This dynamic, phase-based approach allows the algorithm to adapt its detection criteria to the specific characteristics of each fall phase, improving reliability without sacrificing detection sensitivity.
2Speed
If the device detects large acceleration changes to identify falls, then the response speed is improved, but the ability to distinguish falls from non-fall events deteriorates
Solution Approach 1:
The system performs preliminary detection of the free-fall phase before the actual impact occurs. By detecting the characteristic low-acceleration free-fall period first, the system can prepare for imminent impact detection, enabling faster overall response. This preliminary action allows the system to distinguish falls from non-fall events even when both produce similar impact signatures, because only falls exhibit the characteristic free-fall phase preceding the impact.
3Reliability
If the fall detection algorithm is made more sensitive to distinguish falls from non-falls, then the false positive rate is reduced, but the device complexity increases
Solution Approach 1:
The complex fall detection problem is segmented into three manageable phases with distinct detection criteria: free-fall detection (acceleration magnitude below threshold), impact detection (sudden acceleration spike), and post-fall verification (characteristic motion patterns). Each phase has simplified, phase-specific detection rules that are computationally efficient. This segmentation reduces overall algorithm complexity while maintaining high reliability through the cumulative evidence from all phases.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution effectively reduces the number of false positive indications by accurately distinguishing between fall and non-fall events, enhancing the reliability of fall detection systems and minimizing unnecessary emergency alerts.
Implementation Method 1
The accelerometer senses changes in acceleration to facilitate characterization of orientation, movement, and impact associated with a current movement event that may be a fall event
Implementation Method 2
A barometer senses relative pressure of the surroundings and can indicate a device's level relative to the ground by correlating the lower pressure present at higher distances above the ground and higher pressure when closer to the ground
Implementation Method 3
The wireless transmitter can transmit a message to a base station connected to a wired connection, or a wireless connection, for example a long range wireless link such as cellular telephony or cellular modem, or a short range wireless link such as Wi-Fi, Bluetooth, RFID, and the like
Data Source
AI summary
Methods reduce the likelihood of an MPERS device falsely reporting a high acceleration event as a fall impact. The device stores acceleration data acquired before the high acceleration and afterward. If a number of samples of magnitude values from accelerometer sensors in the device acquired during as interval before the high acceleration that approach 0G exceeds a predetermined number, the high acceleration is deemed from a non-fall. Acceleration sensors can also indicate an orientation change before/after the high acceleration, and a barometric pressure sensor can do the same, to further characterize an event as a non-fall. A method compares current event data to composite data sets that have been determined from a plurality of empirically derived data sets of fall and non-fall events. High correlations can indicate falls, or non-falls, respectively. Further statistical analysis of data acquired after an event reduces the likelihood of falsely indicating a fall.